CHAPTER - 5 — a ee a SAMPLING DISTRIBUTIONS ' a 5.4 INTRODUCTION The outcome of a statisti asa descriptive representation, i ipl a dice are tossed and the sum of the numbers on the faces is the outcome of inte ee i i: rd a numerical value. However, if the students of a certain school are given blood tests an he. type of blood is of interest, then a descriptive representation might be most useful, A person’s blood can be classified in 8 ways, It must be AB, A, B or O, with a plus or minus sign, depending on the presence or absence of the Rh antigen. The statistician is primarily concerned with the analysis of numerical data. For the classification of blood types, it may be convenient to use numbers from I to 8 to represent the blood types and then record the appropriate number for each student. In any particular study, the number of possible observations may be small, large but finite, or infinite. For example, in the classification of blood types we can only have as many observations as there are students in the school. The project, therefore results in a finite number of observations. On the other hand, if we could toss a pair of dice indefinitely and record the sums that occur, we would obtain an infinite set of values, each value representing the result ofa single toss of a pair of dice. In this chapter we focus on sampling from distributions or populations and study such .n and sample variance. [JNTU (K) Dec. 2013 (Set No.1)] ical experiment may be recorded cither as a numerical value or important quantities as the sample mea! 5.2 POPULATION AND SAMPLE discussing the notions of population and sample. with which we are concemed, whether this number be finite There was a time when the word population We begin this section by The totality of observations or infinite, constitutes what we call population. There lati referred to observations obtained from statistical studies about people. Today, the statistician thes the term to refer to groups of people, animals, or all possible outcomes from some complicated biological or engineering system or numerical data, Thus the population does not imply living beings alone, ¢.g., we speak of the population of births, heights, weights, prices of Vegetables and so on. Definition : Population (or universe) a subject of investigation. For example, (i) the population of the heights of Indians, (ii) the population of Nationalised Banks in India, ete. 235 is the aggregate or totality of statistical data forming © scanned with OKEN ScannerOL ———— 3s Probably axons polation is tine ty 4 My if population ts denoted by yey, Exam Fehere are G00 students inthe school that we el he, lype, we say at ate bien of size 600, The numbers on the Cards j heights of residents in a certain city, al the © population with finite size, 7 ‘The observations obtained by measuring the almospheric pressure eve ay Past on into the future, or all measurements on the depth of a lake from any cone Position, are examples of population whose sizes are infinite. ia In the field of statistical inference the statistician is interested in Arriving at con Conceming a population when it is impossible or impractical to observe the entire observations that make up the population, : x For example, in attempting to determine the average length of life of a Certain by, Tight bulb, it would be impossible to test all such bulbs if we are to have any left tg sa Therefore, we must depend on a subset of observations from the population jg make inferences conceming the same population. This brings us to consider the "ig, sampling. Definition : The number of observa Population, tt may inite or infinite. Size Sampling : Most of the times, study of entire population may not be oss, carry out and hence a part alone is selected from the given population. A Pottion of Population which is examined with a view to determining the Population charactergig called a sample. i.e., a sample is a subset of population and the number of, Objects iny sempleis called the size of the sample. Size of the sample is denoted by » The process of selection of a sample is called sampling. It is quite often used ing day-to-day practical life, For example, (i) to assess the quality of a bag of Tice, sug, Wheat or any other commodity, we examine only a portion of it by taking a handful; from the bag and then decide to purchase it or not. The portion selected from the bag called a sample, while the whole quantity of rice, sugar or wheat in the bag is the opulato (1) to estimate the proportion of defective articles in a large consignment, only apo Ge., a few of them) is selected and examined. The portion selected is a sample (it) produced in India is the population and the Nano cars is the sample. When information is collected in respect of every individual item, the enquiryiss to be done by Complete Enumeration or Census. For example, during the Censis Population (which is done every ten years in Indi), information in respect of each indivi person residing in India is collected. ‘This method gives information for each and ete unit of the population with greater accuracy. But this method involves multiplicity’ causes viz., administrative and financial implications, time factor, etc. In most ci statistical inquiry, because of limitations of time and cost, only a portion (i.e., a sampl) the available source of information is examined, and the data collected from them. T Process of partial enumeration is known as Sample Survey. The results are then gene? and made applicable to the whole ficld of inquiry. ‘This is known as Sampling Thus in estimating the characteristics of the Population, instead of enue! entire population, only the individuals in the sample are examined. Then the 5 characteristics are utilised to approximately estimate the population, The error inv " 0 ing error and is inherent and unavoidable ing mn) ‘antage of sampling process lies in C™ © scanned with OKEN Scanneree sampling Distributions aay ste population chew universal set forthe sample. The statistical constants like mean, stand -viation, correlation coefficient, etc. obtained for the population are called parameters. Similarly, constants for the sample drawn from the given population i.e., mean (x), standard deviation (S) etc., are called the statistic. If om inferences from the sample to the population are to be valid, we must obtain samples that are representative of the population, Alll too often, we are tempted to choose a sample by sel esting the most convenient members of the population. Such a procedure may ladtoctoneos inferences conceming the population. Any sampling procedure that produces inferences that consistently over estimate (ot) consistently under estimate some characteristics of the population is said to be biased. To eliminate any possiblity of bias in the sampling procedure itis desirable to choose a random sample in the sense that the observations are made independently and at random. 5,3 DIFFERENT METHODS OF SAMPLING Some important methods of sampling are discussed below. |. Probability Sampling Methods 4, Random Sampling (or Probability Sampling) It is the process of drawing a sample from a population ini such a way that each member of the population has an equal chance of being included in the sample. The sample obtained by the process of random sampling is called a random sample. For example : (i) A handy of cards dealt from a well - shuffled pack of cards is @ random sample. (ii) Selecting randomly 20 words from a dictionary is a random sample, (iii) Choosing 10 patients from a hospital in order to test the efficacy of a certain newly - invented drug. If each element of a population may be selected more than once then it is called sampling with replacement whereas ifthe element cannot be selected more than once, it is called sampling without replacement. Note : If Nis the size of a population and m is the sample size, then (i) The number of samples with replacement =N" aC (ii) The numberof samples without replacement 7. 2. Stratified Sampling (or Stratified Random Sampling) “This method is useful when the population is heterogeneous. In this ype of sampling, the population is first sub-divided into several parts (or small groups) called strata according to some relevant characteristics so that each stratum is more or less homogeneous. Each stratum i called a sub-population, Then a small sample (called sub-sample) is selected from each stratum at random. All the sub-samples are combined together to form the stratified sample which represents the population properly. The process of obtaining and examining a stratified sample with a view to estimating the characteristic of the population is known as Stratified Sampling. © scanned with OKEN Scanner238 Probably and guy f . 8 City hay) with a view to studying their economic condition, For this Purpose I city area is divided into a number of strata, according 10 economic condition op lt inhabitants, as measured by anmual income (say). ‘Thus, localities mostly inhabit people with more ot less similar annual income may be included under one stratum, 74 families are then chosen af random from each so that the sum total of all the fami ye all the strata is 500, 2 3, Systematic Sampling (or Quast - Random Sampling) As the name suggests this means forming the sample in some systematic manne taking items at regular intervals. In this method, all the units of the population AFC arrange in some order. If the population size is finite, all the units of the population are arranged i, some order. Then from the first k items, one unit is selected at random. This unit ‘and every kth unit of the serially listed population combined together constitute a systemajg sample. This type of sampling is known as Systematic Sampling. The difference between random sampling and systematic sampling lies in the fag that in the case of a random sample all the members have to be chosen randomly, wheres, in the case of a systematic sample only the first member has to be chosen at random, Il, Non-Probability Sampling Methods. 4. Purposive Sampling (or Judgement Sampling) When the choice of the individual items of a sample entirely depends on the individual judgement of the investigator (or sampler), it is called a Purposive or Judgement Sampling In this method, the members constituting the sample are chosen not according to some definite scientific procedure, but according to convenience and personal choice of the individual, who selects the sample. Two or more such independent purposive samples may give widely different estimates of the same population. In this type, the investigator must have a good deal of experience and a thorough knowledge of the population. Purposive selection is always subject to some kind of bias. This method is suitable when the sample is small. For example, if a sample of 20 students is to be selected from a class of 100 to analyse the extra-curricular activities of the students, the investigator would select 20 students who, in his judgement, would represent the class. 5. Sequential Sampling It consists of a sequence of sample drawn one after another from the population depending on the results of previous samples. If the result of the first sample leads to @ decision which is not acceptable, the lot from which the sample was drawn is rejected: But if the result of the first sample is acceptable, no new sample is drawn. But if the fist sample leads to no clear decision, a second sample is drawn and, as before, if requited,® third sample is drawn to arrive at a final decision to accept or reject the lot. It is widelY used in Statistical Quality Control in factories engaged in mass production and other area 5.4 CLASSIFICATION OF SAMPLES Samples are classified in two ways, ici’ Farge sample : Ifthe size ofthe sample (n) > 30, the sample is sad to be tte sample. For example, let us select a stratified sample of 500 families from a cit ! 50,000 families, © scanned with OKEN Scannersampling Distributions - : . 239 2, Small sample : If the size of the sample (n) < is sai sini 9 Sat Sample, iple (n) < 30, the sample is said to be small In sampling with replacement, each member of the population may be chosen more than once, since the member is replaced in the population. Thus sampling from finite population with replacement can be considered theoretically as sampling from infinite population, Whereas, in sampling without replacement, an element of the population cannot be chosen more than once, as itis not replaced. Therefore the sampling from finite population without replacement can be considered theoretically as sampling from finite population only. Definition : Let X, Xen, ben independent random variables each having the same probability distribution f(x), We thea define X,, Xp. X, to be a random sample of size n from the population f(x) and write its joint probability distribution as Spy Kay eerees Xy) = AO )s MD) serve KE) Cur main purpose in selecting random sample is to elicit information about the unknown population parameters. Statistic ; It is a real valued function of one or more random variables not involving any unknown parameter. ‘Therefore random variable and has a probability or frequency distribution. 5.5 PARAMETERS AND STATISTICS Parameter is a statistical measure based on all the units (or observations) of a population. Statistic (or sample statistic) is a statistical measure based only on all the units selected in a sample. For example, the statistical constants of the population namely mean }, variance 07 are usually referred to as parameters, statistical measures computed from the sample observations alone e.g., sample mean (X), sample variance (6?) etc, are usually referred to as statistics. Tn other words, the mean, median, mode, Standard deviation, variance measures of the parameters and the measures obtained from the sample of the population the random sample. So statistic is a function of statistic is a population are called are called statistics. ‘Thus parameters refer to population while statistics refer to sample. Note: As the units selected in two or more samples drawn from a population are not the same, the value of a statistic varies from sample to sample, but the parameter always peony constant (since all the units ina population remain the same). ‘This variation in the value of a statistic is called sampling fluctuation. ‘A parameter has no sampling fluctuation. Usually, statistic is used to estimate the value of an unknown parameter obtained from the sample, is a function of the sample values only. 1. The Sample Mean : Definition : LX, Xy mean is defined by the statistic 1X, represent a random sample of size n, then the sample © scanned with OKEN Scanner240 Probabllty and gy, Mey an Note that the statistic X has the value X= 2 3 When x assumes a it values, . For instance, the term sample mean is applied to both the statistic x ‘ang ‘iy computed value X. 2. The Sample Variance : Definition : If.X,, X, represent a random of size m, then the sample vagy, P_nDxP-ExP is defined by the statistic s? yee Gey (measure of variability of data about the mean) Note that s*is essentially defined to be the average of the squares of the deviations of iy observations from their mean, With the change that the sum of the squared deviations ig divided by m — 1 and not n. Ex: A comparison of coffee prices at 4 randomly selected grocery stores in a city shoved increases from the previous month of 12, 15, 17 and 20 rupees for 1 kg. Find the variance of this sample of price increases. Solution: Calculating the sample mean, we get 1241541742 ge BAHN HITH20 gp. 4 se (12 = 16)? + (15-16)? + (17-16)? + (20-16)? 3 y? +? + (4)? 3 3. The Sample Standard Deviation : The sample S. D., denoted by s , is the positive square root of the sample variance. Note : In the above example $.D., s = 3473 . 5.6 SAMPLING DISTRIBUTION Sampling theory is the study of relationships between a population and samples drawn from the population, and it is applicable to random samples only. We need to obteit maximum information about the population with the help of samples, i.e, to determine th true value of the population parameters (population mean, population S. D., populace proportion, etc.) by using sample statistics like sample mean, sample S. D., sam proportion, etc., and to find the limits of accuracy of estimates based on samples. It helS us to determine whether the differences between two samples are actually due to cha" variation or whether they are really significant. Sampling theory also useful in testi" ° hypothesis and significance which is important in the theory of decisions. (4¥ + ad © scanned with OKEN Scannergempling Distributions 241 5,7 SAMPLING DISTRIBUTION OF A STATISTIC Sampling distribution of a statistic j it cc rions values of eats fatistic is the frequency distribution which is formed sahvarious value of istic computed from different samples of the same size drawn fom a ame Pu lation. We can draw a large number of samples of same size from a porulat : xed size, each sample containing different population members. Any statistic (statistical measure of sample) like mean, median, variance, etc. may be computed for each of these samples. As a result a series of various values of that statistic may be obtained. These various values can be arranged into a frequency distribution table, which is known as the sampling distribution of the statistic. Sampling may be done with replacement or without replacement. Sampling with replacement means that the same unit of the population may be included in each sample more than once, Sampling without replacement means that the same unit of population may not be included in each sample more than once. Let us consider a finite population of size N and let us draw all pos: N n\(N-n)! Compute a statistic ¢ (such as sample mean, $.D., etc) for each of these samples. The be the values of le random samples each of the same size n. Then we get KGa =k (say) samples. value of a statistic ¢ may vary from sample to sample. Let f,,ty)----f statistic k for the k samples. Each of these values occur with a definite probability. Thus we can construct a table showing the set of values ),/,,...t, of ¢ with their respective probabilities. This probability distribution of r is known as the Sampling Distribution of ¢ Thus sampling distribution describes how a statistic ¢will vary from one sample to the other of the same size. Although all the & samples are drawn from the given population, the members included in different samples are different. IfNis large, then the number & of all possible samples is also large (ie., sampling the values t,,fys-ty Of ¢ may be arranged in the ion and the limiting form of this relative frequency distribution when k > is called the sampling distribution of statistic ¢. A statistic (ie., sample mean, sample S.D, etc.) has always a sampling distribution, but a parameter (ce. population mean, population SD.) has no sampling distribution. without replacement). In such a case, form of a relative frequency distributi If the statistic ¢ is mean, then the corresponding distribution of the statistic is known as sampling distribution of means. regarded as a random variable ¥ and each sample mean Tue of this new random variable ¥. Let these values be: .¥, can be used to form a frequency ‘The sample mean can be then constitute as the observed va ‘These mean values ¥, %2,%» distribution. Then this frequency distribution of the statistic x is known as the sampling distribution of the sample mean. Similarly the sampling distribution of standard deviation or variance may be constructed with the various values of standard deviation or variance tespectively. BR Ba Fy © scanned with OKEN Scanner242 Probability ang , ; at bution ofa statistic is tha it ap, > appre, The main characteristic of the Sampling dis shy is sufficiently lange (greater than 30), Another important feature of the sampling alae sin cs thatthe mean and the standard deviation ofthe sampling distribution of sim, bear a definite relation to the corresponding parameters ic., mean and standard dey parent population. These characteristics of the sampling distribution help us: (i To estimate the unknown population parameter from the known Statistic i To set the confidence limits of the parameter within which the parameter va expected to lie. (ii) To test a hypothesis and to draw a statistical inference from it. 5.8 CENTRAL LIMIT THEOREM normal distribution even when the population distribution is not normal provided the Sam, sta ation op ues ane If ¥ be the mean of a random sample of size n drawn from a population " havin mean and S.D. o, then the sampling distribution of the sample mean ¥ is approximate imately anormal distribution with mean and $.D, = S.E, of ¥=-2 provided the sample size, vn is large (1230). This is established by central limit theorem stated below (without proof Theorem : If X be the mean of a sample size _n drawn from a population with mean Hand S. D. o then the standardized sample mean E-u olvn is a random variable whose distribution function approaches that of the standard norm! distribution N(z;0,1) as n—>00. 5.9 STANDARD ERROR (S. E.) OF A STATISTIC The standard error ofa statistic ¢ (i.e., S. E. of sample mean, or sample S.D.) is tie standard deviation of the sampling distribution of the statistic. Thus S.B. of sample mean ¥ is the $.D. of the sampling distribution of sample mean. It is used for assessing the difference between the expected value and observed value. It plays an important role in large sample theory and forms the basis in tests of hypothesis or tests of significance. I gives an idea about the reliability and precision of a sample. S. E. enables us to Seer the confidence limits within which the parameters are expected to lie, For example," probable limits for population proportion P are given by p+3./pq/n. . W The standard errors (S.E.) of some of the well-known statistics are given bel ation (without proof), where } the population mean, o? the population variance, P the pop"! P the proportion, » is the sample size, ¥ the sample mean, s? the sample variance and sample proportion. Note : S.E. of a statistic may be reduced by increasing sample size "™ results in corresponding increase in cost, time and labour, etc. put th a >| © scanned with OKEN Scannersampling Distributions - Formulae for 1. S.E. of sample mean ¥ s o + It is written as S. E. @) ir nas @) Tr PO | where Q=1-P n V2n 4. S.E. of the difference of two sample means % and % 7 Ley SE. of (G-%)= J+ where ¥ and %, are the means of two random nm samples of sizes n, and n, drawn from two populations with $.D. o,and o, respectively. 5. S.E. of (pj, —p2)= RO, BOs where ‘p, and p, are the proportions of two mom random samples of sizes n, and 7, drawn from two populations with proportions 2. S.E. of sample proportion 3. S.E. of sample S. D (s) R and P, respectively. 2 Ge ofS ee 6. S.E. of (s,-52)= 2m, * 2m For a finite population of size N, when a sample is drawn without replacement, we have o [N-n (0 S.E. of sample mean =——= No : PQ [N=n (ii) S.E. of sample proportion - 2. ea n when the sample is drawn without replacement, Formulae For an infinite populatior 1 and 2 remain the same. Proposition (Statement without proof): The mean and S. E. of sample mean ¥ | mean and §.D. of the sampling distribution of =) when samples of the same size n ate drawn from a population having mean # and S.D. are given by Mean of ¥ = E(%)=1t and SE. of ¥ = : 5.10 SAMPLING DISTRIBUTION OF MEAN (co KNOWN) The probability distribution of X is called the sampling distribution of means. ‘The sampling distribution of a statistic depends on the size of the population, the size of the samples, and the method of choosing the samples, © scanned with OKEN ScannerON 244 Probabilly and stay ly Let XX yyy be the w random samples drawn from & population of, N with mean 41 and variance ¢? and X is the mean of samples. Infinite Population : Suppose the samples are drawn from an ing Nite population i.e.,N >. (or) sampling is done with replacement, then (4) The mean of the sampling distribution of means, uy =H i.e. ER) =p Xy+Xp+Xy+tX 7 Proof: We have X = Since X isa linear combination of the random variables X1,X2,X3,.yXq5 therefore itis also a random variable, . Mean of X= E(X) = Jt Ecx, +X +X3+..+X,) n " sin (EQ) + BX.) +E (Xs) +... EX, )] Gi) Variance, o3 S.D of mean, o; Proof: Ifthe sample is drawn with replacement, then X,,X,X3,...X,y are independent x ng? o random variables, Thus we have Var (X)= "0 = &— ® ‘The sampling distribution of ¥ will be approximately normal with mean and 2 variance n provided that the sample size is large. Standardized sample mean, z=*—!_ olvn Finite Population : Consider a fi nite population of size N wi Draw all possible s ; jon & t ith mean jt and standard deviatiot ‘amples of size n without Te] Placement, from this population. TH he, ER)=p © scanned with OKEN Scannering Distributions sampling an sy (if the sample is drawn without replacement) and N-n Here, the factor ( No "), often called the finite population correction factor. We note that this term tends to become closer and closer to unity as population size becomes larger and larger. Normal Population (Small Sample) : Sampling distribution of X is normally distributed even for small samples of sige n < 30 provided sampling is from normal population. Non- Normal Population (Large Sample) : Consider a population with unknown (non-normal) distribution. Let the population mean jt and population variance o be both finite. Let the population be finite or infinite. In case the population is finite assume that the population size N is atleast twice the sample size n. Draw all possible samples of size m. Then the sampling distribution of X is approximately normally distributed with mean j1, = 4 and variance 2 o,2 = & provided the sample size is large (i.e., 2 30). n 5.11 SAMPLING DISTRIBUTION OF PROPORTIONS Let p be the probability of occurrence of an event (called its success) and q=1~p is the probability of non-occurrence (called its failure). Draw all possible Samples of size m from an infinite population. Compute the proportion P of success foreach of these samples. Then the mean Hp and variance op’ of the sampling distribution of proportions are given by distributed, the sampling distribttion of proportion large. For finite population (with replacement) while population is binomially is normally distributed whenever 7 i of size N, we have N-n epee a) ISTRIBUTION OF DIFFERENCES AND SUMS 5.12 SAMPLING DI Let pnd 0, be the mean end standard deviation of sampling distribution of ned by computing 5, for all possible samples of size m, drawn from and Gg, be the mean and standard deviation of sampling ‘obtained by computing S, for all possible samples of size n. a statistic $, obtai population 4. Also let Hs2 distribution of statistic 5, nother different population B. drawn from a1 © scanned with OKEN Scanner| 246 Probably and gy, ht, the difference of the statistic from, Now compute the statistic S, - S from the two populations A and B, possible combinations of these sampl Then the mean pg, ¢, and the standard deviation Og). 5; Of the sam, distribution of differences are given by en aareee2 Hoyos = Hay 7 Hsp and Gs,_ 5: = Ys, +s, assuming that the samples are independent. ty Ml he Ping Sampling distribution of sum of statistics has mean ts, .», nd standard devi Os, - «: given by . feqarv=na| Hey ssa = Hoy + Hey Ad 5,452 = Ys, +05, For example, for infinite population the sampling distribution of sums Of means has mean Hs,,s and Fg,,x) given by Hysx, = Ha, + Ha, = Hy +H, and 7? lo2 , o, Sum = You toy? = te 2 m+ Ox ene For sampling distribution of differences of proportions, we have bp, Ps = Mp, Hp, = Py P2 and 7 N , P24: and on-n,= fovoy = fh +B& What is the value of correction factor ifn = 5 and N = 200. Solution: Given N = Size of the finite population = 200 n = Size of the sample = 5 - N=n _ 200-5 _ 195 . Correction factor =~ = 39927 7 199 7 [SEEDER Find the value of the finite population correction factor for n = 10 and N= 100. [INTU 1999, 2000S, (A) Dec. 2009, Nov. 2010, Apr. 2012 (Set No.3)] Solution: Given N = Size of the finite population = 1000 n ~ Size of the sample = 10 990 z, Correction factor DENEOA)] How many different samples of size two can be chosen, from finit population of size 25. Solution: We can take NC, samples of size n from the population of sizé N Here N= 25, n=2 ae? +. We can take 5, = 300 samples of size 2 from finite population of Si#° ~ a © scanned with OKEN Scannerrr ling Distributions om 247 Apopulation consists of five numbers 2, 3, 6, 8 and 11, Consider all possible samples of size two whi 6 this population. Find Wwo which can be drawn with replacement from (a) The mean of the population, (b) The standard deviation of the population. (c) The mean of the sampling distribution of means and (d) The standard deviation of the sampli istributi i. standard error of mean), ipling distribution of means (i.e., the [JNTU Nov 2004, April 2005 (Sets 3, 4), (A) Dec. 2009 (Set No. 4)] Solution : (a) Mean of the population is given by 24+34+6+8+11 0 3 — (») Variance of the population (a) is given by n (2-6)? + (3-6)? + (6-6) + (8-6)? + (11-6) 5 16+94+0+4425 = SS 708 2. 6 = VOB ie, 6 = 3.29 () Sampling with replacement (Infinite population) : The total no. of samples with replacement is Ne = 52 = 25 samples of size 2 Here N= population size and » = sample size listing all possible samples of size 2 from population 2, 3, 6, 8, 11 with replacement we get 25 samples (2,2) (2,3) (2,6) (2,8) 2,11) 3,2) (3,3) (3,6) (3,8) GAL) 2) (6,3) 6) 8) 1) (82) (8,3) (6) 8) BI) (11,2) (11,3) a6) (8) 11) Now compute the arithmetic mean for each of these 25 samples. The set of 25 means X of these 25 samples, gives rise to the distribution of means of the samples. known as sampling distribution of means. ‘The samples means are 25 A i 4 5 6.5 2.5 3 45 55 7 A 4.5 6 7.0 8.5 @ 5 5.5 1 8 9.5 as 7 8.5 9.5 ul and the mean of sampling distribution of means is the mean of these 25 means, ~~ s © scanned with OKEN Scanner248 (4d) population) (c) (4) Probabity yy ’ Shay, Sum of all sample means in (1) 4 We GF Mlustrating that jug = 1. The variance o,7of the sampling distribution of means is obtained by sub the mean 6 from each number in (I) and squaring the result, addin’ tt members thus obtained, and dividing by 25, 8 all 95 i= _ 135 5 iy = Js = 5-40 and thus o, = 540 = 2.33 Clearly, for finite population involving sampling with replacement (or in Mite ke oro, = z 2 a222 7 mn ~ 2 Solve the above example (i.e. Ex. 4) without replacement, Solution: (a) u=6 (6) 6 =3.29 Sampling without replacement (finite population) : The total no.of samples without replacement is “C, = $C, = 10 samples of size? The 10 samples are (2, 3) (2, 6) (2,8) (2, 11) 3, 6) 3,8) 3,1) (6, 8) (6, 11) (8, 11) The selection (2, 3) is considered same as (3, 2) The corresponding sample means are 25 4 5 65 45 55 7 7 8.5 9.5 The mean of the sampling distribution of means is (2.5444+546544.545.5+7 47485495) ag Mlustrating that py = The variance of sampling distributions of means (25-6) + (4-6)? 4 neet(9.5~ 6) I 10 sate) 2 = 4.05 and oo; = 2.01 e Showing that o,? = = (2 n for sampling without replacement. 4 © scanned with OKEN Scannergampling Distributions wi A population consists of 5,10,14,18,13,24. Consider al possible samples of size two which can be drawn without replacement from the population. Find (a) The mean of the population (0) The standard deviation of the population (c) The mean of the sampling distribution of means (d) The standard deviation of sampling distribution of means. [NTU (HH) Nov. 2010 (Set No.3)] Solution: (a) The mean of the population p is given by Bx _S+10 +14 418413424 n 6 ~ (b) Variance of the population g? is given by Xxx)? ee 2Gic) n 1 = HG=14)* +(10-14)? + (04-14) +18 14)? + 13-14)? + 24-14) =)p61+16+0+16+14 100) = 24 3567 (c)All possible samples of size two (the no. of samples = °C, =15) and their means are shown in the following table. ‘Sample No. ‘Sample values Totalof ‘Sample mean Sample values 1 5,10 15 1S 2 514 19 95 3 518 2B 1s 4 5,13 18 9 5 524 29 145 6 10,14 24 2 7 10,18 28 4 8 10,13 B us 9 10,24 34 7 10 14,18 32 16 ul 1413 2 BS 2 14,24 38 19 3 18,13 31 155 4 18,24 42 a1 15 13,24 Eu 185° Total 210 © scanned with OKEN Scanner250 Probability ang ig 2. Mean of sample means = 2 =14 ie., The Mean of the sampling distribution of means is p= 14 Mlustrating that py = (d) The variance of sampling dstribution of means a Rua s-147 +(95-14)? +1519 +019" (14.514)? + ot 21-14)? #18 5-14)"] = Rts2.25+20 25 +6.25+2540.25+4+0+6.25 $94440.25+25+2.25 +494 20.25] 14.2666 1S ©. Standard deviation of sampling distribution of means is o, = V14.2666 = 3.78 PEPE] Let u, = (3. 7. 8). uy = (2, 4). Find (@) 4, ) (c)_ Mean of the sampling distribution of the difference of means y,,. @) os, (©) OF, (A the standard deviation of the sampling distribution of the diffrences of meas [JNTU Nov. 2008 (Set No.3) © o, = 2-3 +B-9" Ly 2 © scanned with OKEN Scanneristributi enping Distributions ost OG = DA SEDE HOF HH O= TP HCD aed 6 - ff 6 V3 Find the mean and Standard deviation of sampling distribution of variances for he population 2.3,4,5 by drawing samples of size two (a) with replacement (4) without replacement. [JNTU (A) 2009, Nov. 2011 (Set No. 2,4), (K) II Sem. June 2015 (Set No. 4) Solution: Population size, N = 4, sample size, n = 2. (a) Number of samples of size two with replacement is N" = 4? = 16. They are (2,2), (2.3), (2,4), (2,5). (3,2), 3,3), (3,4), (3,5)s (4.2). (4,3), (44)s (4,5), (5,2), (5,3), (5,4), (5,5). We now compute the statistic variance for each of these 16 samples. Variance for the sample (22) with mean 2 is 412-2)" +(2-2)"]=0 Similarly, the variance for sample (2,3) with mean 2.5 is. A225 +3-25)1= 028, ‘Thus the variances of these 16 samples are 0 0,25 1 2.25 | 0.25 0 0.25 L ! 025° «0 0.25 | 225 1 025 (0 ‘Thus the sampling distribution of variances (with replacement) is g | 0 025 | 1 | 225 Frequency | 4 6 4 | 2 4(0) +6(0.25) + 1(4) + 22.25) _ 16 Mean of S.D. of variances = Variance of S.D. of variances = fa(o-0.625)? + 6(0.25- 0.625)? + 4(1-0.625)* +2(2.25-0.625)] 16 =e 16 (b) This is left as an exercise to the reader. = 0.5156 © scanned with OKEN Scanner252 Probability ang [ERNETTEUG] A population consists of six numbers 4,8,12,16,20,24. Consider aj Sa, samples of size two that can be drawn without replacement from this popuation, ping Me a) The population mean b) The population standard deviation ©) The mean of the sampling distribution of means ) The standard deviation of the sampling distribution of means |JNTU 2000, (H) Nov. 2009 (Set Nay Solution: a) The mean of the population, 4 4+ 8412416420424 6 =hin4 6 b) The variance of the population, o? = she -x? 34-149 + (8-14)? + (12-14)? + (16-14)? + (20-14)? + (24-14)?] 1 = 7100 +3644 +4436+ 100] = 7246.7, . The population standard deviation, = (46.67 = 6.83 The number of samples that can be drawn from the population of size 6 without 6 replacement is °C, i.e, 374) oF 15. The (15 samples are) sampling distribution is, {eo (4,12), (4,16), (4,20), (4,24), (8,12), (8,16), (8,20), (8,24), (12,16), (12,20), (12,24), (16,20), (16,24), (20,24) } The means are 6, 8, 10, 12, 14, 10, 12, 14, 16, 14, 16, 18, 18, 20 22 c) The mean of the sampling distribution of means is 1 Hy = Fg(6+8+10412+14+10+12 4144164144164 18+18+20+22] = 05 14 15 eet © scanned with OKEN Scannerampling Distributions (253 ) The variance of the sampling distribution of means is G2 4G -xP 16-14)? = gllé 14) + (8-14)? + (10-14)? + (12-14)? 414-14)? + (10-14)? +(12~14)? (14-14)? 4 (16-149? + (14-14)? 416-14)? + 08-149? + (18-14) +(20-14)? +(22-14)°] 1 jglO4+36+16 +440416+44044+044416+16+36+ 64] Hence the standard deviation of the sampling distribution of means is 0, = Vi8.67 =4.32V2- Note : p=pz and a? #03 Samples of size 2 are taken from the population 1,2,3,4,5,6 (i) with eplacement and (ii) without replacement. Find a) The mean of the population b) Standard deviation of population ) The mean of the sampling distribution of means d) The standard deviation of the sampling distribution of means. Jerify that means of sampling distribution is equal to the mean of population and standard ieviations of the means of sampling distril [INTU (H) Nov. 2009,(A) Nov. 2011 (Set No.3) (K) June 2015 (Set No, DI bution are not equal to the standard deviation of the sopulation. Sol - _ 14243444546 214 5 (o (a) The mean ofthe population, H=——— gg n is at D 5-H (b) The variance of the populatio AG 3,5) +23. +6-3.5)° + (4-35)? 45-35)? +(6-3.5)") [6.25+2.25+0.25+0.25+2.25+6.25) 1 6 _ The standard deviation ofthe population is, o= 2.917 =1.71. 2 36. (c) Number of samples of size two with replacement is N" = © scanned with OKEN ScanneroN 254 Probabllity ang ay They are (yt) (2) (3) 44) (1,5) (1,6) 1) 22) 23) @4) (5) (2,6) G1) G2) G3) G4) G5) G6) (4.1) 4.2) 43) 44) (45) (4,6) (5.1) 2) 3) (5,4) (5,5) (5,6) (6.1) 2) (6,3) (6,4) (6,5) (6,6) :. The number of samples, 1 = 36 Their corresponding means are 1, 15, 2, 25, 3, 3.5 15, 2, 2.5, 3, 3.5, 4 2, 25, 3, 35, 4, 45 25, 3, 3.5, 4, 45, 5 3, 35, 4, 45, 5, 55 35, 4, 45, 5, 5.5, 6 The mean of the sampling distribution of means is 1 My = gg Ul 41542+254343.54254..46] = 26 35° 36 (@) The standard deviation of the sampling distribution of means is x ; DGi-vsy a n [(-3.5)? + (1.5-3.5)? + (2-3.5)? +(2.5-3.5)? +3-3.5) EA 36 +(3.5-3.5) +(1.5-3.5)? +....4(6-3.5)7] Itis clear that the mean of the sampling distribution of means is equal to the mean of population and standard deviation of the means of sampling distribution is not equal !° standard deviation of the population. (ii) This is left as an exercise to the reader, ik GETS] Samples of size 2 are taken from the population 3,6,9,1527 ™ replacement. Find a) The mean of the population 4) The standard deviation of the population ¢) Mean of the sampling distribution of means d) The standard deviation of the sampling distribution of means: a [SNTU GH) Now. 2009 62 | © scanned with OKEN Scanneryr ganpling Distributions ; oss 34649415427 _ 60 _ 5 “s (p) Standard deviation of the population, = LG, -ny? Yn y+ ( 12 solution: (a) Mean of the population, #= 2)? +(9-12)? +(1 5 i = |481+36+9+94225) =, [2 5 3 (c) The sampling distribution with replacement is G3) G6) G9) GIS) G27) (6,3) (6,6) (6,9) (6,15) (6,27) (9,3) (9,6) (9,9) (9,15) (9,27) (15,3) (15,6) (15,9) (15,15) (15,27) (27,3) (27,6) (27,9) (27,15) (27,27) The means are 3 45 6 9 1s 45 6 75 105 16.5 6 aS 9 12 18 9 10.5 12 15 21 15 16.5 18 21 27 ‘The mean of the sampling distribution of means, 344.546494154+4.5+..418421427 300 * 25 * as 4853 E =12, (d) The standard deviation of the sampling distribution of means [Rie-1ay +(45- 137 (6-12) 40-12) + (15-12) bs (18-12)? + (21-12? + (27-12)") [1 1g1456.25+36+9+9-+56.25+36+20.25 + 2.25 4-20.25+9+20.25+ 5 940436494 2.254049+81+9+56.25 +36 +814 225] 5.03 If the population is 3, 6, 9, 15, 27 (a) List all possible samples of size 3 that can be taken without replacement from the finite population. (&) Calculate the mean of each ofthe sampling distribution of means. — roam © scanned with OKEN ScanneroN 256 Probability ang ay (©) Find the standard deviation of sampling distribution of means. [NTU Nov 2004, (A) Dec. 2009, (K) II Sem. June 2018 (S015 (or) A population consists of five numbers 3,6,9,15,27. Consider all possible say, of size three that can be drawn without replacement from this population, Fing (i) The population mean. (ii) The population standard deviation. (iii) The mean of the sampling distribution of he means . (iv) The standard deviation of the sampling distribution of the means. IINTU (K) Dee. 2015 (Set yo, yy 34649415427 Solution: Mean of the population, j1 = : es rene) Standard deviation of the population, 2)? + (6-12)? + (9-12)? + (15-12)? +(2 3 S1+364949% B60 FS 8.4853 (a) Sampling without replacement (finite population) : The total number of samples without replacement is “C, =5C, = 10 The 10 samples are (3, 6 9, (3, 6, 15), (3, 9 15), (3, 6 27), (3, 9, 27), (3, 15, 27), (6, 9, 15) (6, 9, 27), (6, 15, 27), (9, 15, 27). Computations : Samplemean| 6 | 8 | 12 | 9 | 13] 15] 10] 14] 16 | 17 x, 6] 8 9 10] 12] 13} 14] 16] 15 | 17 tS 1 1 1 1 1 1 1 1 1 1 (6) Mean of the sampling distribution of means is = SHB 9+ 10+ 124 IS4 4H 5416417 _— 120 a 10 ares (c) sl 12)? + (8-12)? + (9-12)? + (10-12? +12 12? 120 + (13-12) +04-12)°+ 05-12) + 6-12)? 407-1993] 784 = VI33 = 3.651 Let S= {1, 5, 6, 8}, find the probability distribution of the sample ™ = | for random sample of size 2 drawn without replacement. [INTU (K) Nov. 2011 (Set © scanned with OKEN Scannerjampling Distributions ind 257 (OR) A population consists of the four numbers 1, 5, 6, 8. Consider all possible samples ofsize two that can be drawn without replacement from this population. Find (i) The population mean (fi) The population standard deviation (iii) The mean of the sampling distribution of means, (iv) The standard deviation of the sampling distribution of means. [JNTU (K) Nov. 2012 (Set No.1,2)] Solution: Let S = {I,5,6,8} (a) Mean of the population, w= 1t5+6+8 _ 20 _ 4 4 . 1 (b) 8. D. of the population, 6= | 9%) - 0)? 1-5)? (56-3)? + (6-52 +8 -5)? (c) We have S= {1, 5, 6, 8} Here size 2 is to be drawn (1,5) (6) (48) (5,6) (5,8) (6,8) ie., 6 samples Sampling distribution of means are 3, 3.5, 4.5, 5.5, 6.5, 7 . _ _ B4BSH4.545546547 _ 30 |. Mean of sampling distribution, ¥ =~ =“ =5 (d) §.D. of sampling distribution of means, (3-5) +G.5-5)° (45-5)? +(5.5-5) + (65-5)? +(7-5) 3 _ FeRIST02S+ 02542208 E Sei ‘The variance of a population is 2. The size of the sample collected from the population is 169. ‘What is the standard error of mean. n =The size of the sample = 169 4g f a : ratio® [EEETVSTT The mean height of students in a college is 155 ems and standard devi : is 15. What is the probability that the mean height of 36 students is less shat 157 cms. [NTU (A) Nov. 2010 (Set _—— © scanned with OKEN Scannering. Distributions. 259 Solution: }! = Mean of the population = Mean height of students of a college = 155 cm o = S.D of population = 15 cms n sample size = 36 x1 1 Mean of sample = 157 oms = Now z aha 157-155 12 f = 155k anisms P(x < 157) =P (z<0.8) =0.5+ P(0 52.9) Fou | $2.9-51.4 .zos = ye 7176 ©. P(¥ > $2.9) = Plz > 1.76) 0.5 — P@ n= 42 (nearly) ‘The guaranteed average life of a certain type of electric bulbs is 1500 hrs with a $.D of 120 hrs. It is decided to sample the output so as to ensure that 95% of bulbs do not fall short of the guaranteed average by more than 2%, What will be the minimum sample size? {JNTU (A) Nov. 2010, (K) June 2015 (Set No. 3 Solution: Let n be the size of the sample. The guaranteed mean is 1500 sample to be less than 2% of (1500) i.e., 30 hrs. ‘We do not want the mean of the So 1500 - 30 = 1470 ¥ > 1470 1470-1500) 120/Vn normal curve to the left of From the given condition, the area of the probabi nia should be 0.95 ‘The area between 0 and Ania is 0.45 We do not want to know about the bulbs which have life above the guaranteed life. © scanned with OKEN ScannerPp | 264 Probability ang Stati When the area is 0.45, 2 = 1.65 ~ In SP 16S ine Vn = 6.6 1 n= 44 Determine the expected number of random samples having hq, means(a) Between 22.39 and 22.41 (6) Greater than 22.42 (c) Less than 29 37 (A) Less than 22.38 or more than 22.41 for the following data. N = Size of the population = 1500 n = Size of the sample = 36, Number of samples (N,) = 300 © = Population $.D = 0.48, 4 = Population mean = 22.4 Solution : (a) P (22.39 < % < 22.41) = P(-1.26 22.42) = P(z > 2.53) = 0.00057 . Expected no. of samples = 300 (0.00057) = 2 (ec) P(® > 22.37) = P(e < - 3.8) = 0.0001 Expected no. of samples = (300) (0.0001) = 0 (a) P (( < 22.38) and (¥ > 22.41)] = P(e <-2.53 and z > 1.26) = 0.0057 + 0.1038 = 0.1095 “. Expected no. of samples = (300)(0.1095) = 33 The mean voltage of a battery is 15 and S.D is 0.2. Find the probability that four such batteries connected in series will have a combined voltage of 60.8 or more Vol [INTU (A) Nov. 2010 (Set No. 2), (KK) May 20108 (Set No! Sol Let mean voltage of batteries A, B, C, D be X.. X., Xa Xp MH : |, B,C, Dbe X,, Xp Xo Xo mean of the series of the four batteries connected is ae ° BRasSpeKcrRy = Mx, + Hey + xe + Hep = 1S+154+15+15=60 Prrarero = loro, teteos = facoap =04 Let X be the combined voltage of the series _ © scanned with OKEN Scannera sampling Distributions . a When x = 60.8, z= ~—# 3 Then probability that the combined voltage is more than 60.8 is given by P(X 2 60.8) = P(z > 2) = 0.5 - 0.4772 = 0.0228. emo Three masses are measured as 62.34, 20.48, 35.97 kgs with S.D 0.54, 021, 0.46 kgs. Find the mean and S.D of the sum of the masses. Solution : Let the three masses measured for 4, B, C be X,, Xp Xe The mean of the sum of the masses is HX,q+XptXe = HR, + Hxg + HR = 62.34 + 20.48 + 35.97 = 118.79 and 6, pee = yo, toh 40% = ¥(0.54)? +(0.21)* + (0.46) = 0.74 The mean life time of light bulbs produced by company is 1500 hours and $.D of 150 hours. Find the probability that lighting will take place for (a) atleast 5000 h. (8) at most 4200h if three bulbs are connected such that when one bulb burns out, another bulb will go on, Assume that life times are normally distributed. Solution : Let the mean life time of light bulbs Z,, 2, and L, be X, , X,. Then mean of these light bulbs is HX, +X, +%.5 = Hx, + Hx, + HEL; 71500 + 1500 + 1500 = 4500 and $.D is Oustysts = ¥Ot, +t, +13 = f3 as0) = 260. (a) The probability that lighting will take place at least $000 A is xoe 5 5000 ~ 4500) P(X > 5000)= P(e > ~~) = PL? ——~“Geq J = P(E > 1.92) = 0.5 — 0.4726 = 0.0274 (b) ‘The probability that the lighting will take place at most 4200 A is 4200-4500 P(X < 4200) = P (=< esse ) =P (¢<- 1.15) = 0.5 - 0.3749 = 0.1251 ME Nea © scanned with OKEN Scanner266 is produced by company B will be (a) at least 600 N more than (6) at least 450 N more ta Probability ang Say li Determine the probability that the mean breaking strength of g th, cables produced by company A, if 100 cables of brand A and 50 cables of brand B are tot Company | Mean breaking) —$.D Sample Strength Size A 4000 N 300 N 100 B 4500 N 200 N 50 of 0.003 inch. The inner diameter of bearin; S.D of 0.002 inch. @ (ii) @ (ii) NTU (K) Nov. 2011 (SceN4y Given X, = 4000, X, = 4500, c, = 300, o, = 200 and n,= 100 , n= 50 UXg-X, = Hxy— HX, = 4500-4000 = 500 N 7S S54 (200)? , (300)? _ Hy ty = ye SS = Vi700 = 41.2 ‘ Me os 50” 100 0 3 PEATTEEATG] The diameter of motor shafts in a lot has a mean of 0.249 inch and a SD 18s in another lot have a mean of 0.255 inch anda and What are the mean and the S.D of the clearances between shafts and bea: rings selected from those lots ? Ifa shaft and a bearing are selected at random, what is the probability that the shaft will not fit inside, the bearing ? Assume that both dimensions are normally distributed. [JNTU April 2003] Sol ym Let X, be the mean diameter of bearing and X, be the mean diameter of shat. Given X, = 0.255, X, = 0.249 and o, = 0.002, o, = 0.003 Let X, be the mean diameter of the difference in the two diameters. X= X,- X= ,- pw, = 0.255 — 0.249 = 0.006 oy= orto? = (0.002)? + (0.003? = 0.0036 Shaft will not fit inside bearing if d <0, 00.006 _ 0.0036 ~~ 1-664 Probability that shaft will not fit inside bearing = Now P(d <0) = Pe <- 1.66) = 0.5 ~ 0.4515 = 0.0485. © scanned with OKEN Scanneramping Distributions oat REVIEW QUESTIONS 1, Define population, sample and sampling, Give an example each. [JNTU (K) Dec. 2013 (Set No.1) si 2, Define parameter and statistic. 3, Explain sampling distribution and sampling distribution of a statistic. EXERCISE). 1. Find the values of the finite population correction factor for (a)n = 5 and N= 200 (6) n = 50 and N = 300 2. How many different samples of size m= 2 can be chosen from a finite population of size N= 25. 3, A sample of size 400 is taken from a population whose S.D is 16, Find the standard error. 4, The mean weekly wages of workers are with S.D of rupees 4. A sample of 625 is selected, find the standard error of the mean. 5. When we draw a sample from an infinite population, what happens to the standard error of the mean if the sample size is (a) increased from 50 to 200 (b) Decreased from 640 to 40 6. (i) A random sample of size 81 is taken from an infinite population having the mean 65 and S.D 10, What is the probability that X will lie between 66 and 68? (ii) Determine the probability that X will be between 22.39 and 22.41 if a random sample of size 36 is taken from an infinite population having the mean p=22.4 and o=0.048. [JNTU (K) March 2014 (Set No.2)] (i1i)Determine the probability that X will be between 66.8 and 68.3 if a random sample of size 25 is taken from an infinite population having the mean n=68 and o=3- [JNTU (K) March 2014 (Set No. 3)] (iv) Find PCH > 66.75) if a random sample of size 36 is drawn from an infinite population with mean p=63 and ¢=9. _[JNTU(K)March 2014 (SetNo. 4)] 7. Find the probabilities that a random variable having the standard normal distribution will take on a value. (a) between 0.87 and 1.28 (b) between ~ 0.34 and 0.62 [JNTU 1999) 8. The average cost of a studio condominium in the cedar lakes development is Rs, 62,000 and the §.D is Rs. 4200. What is the probability that a condominium in this development will cost at least Rs. 65,000. % Determine the probability that the sample mean area covered by the sample of 40 of | litre paint boxes will be between $10 and 520 square feet, given that a 1 litre of such paint box covers on the average 513.3 square feet with S.D of 31.5 sft. (INTU April 2008 (Set Noy © scanned with OKEN Scanner10, Caleulate the probability that a random sample of 16 computers will faye average life of less than 775 hours assuming that length of life of computer approximately normally distributed with mean 800 hours and S.D 40 hours 11.) Construct S.D of means for the population 3, 7, Hy 15 by drawing sain, of size two with replacement, Determine (a) # (b) 6 (€) SDM (a) ,! a o,. Verity the results. (i) population consists of the four numbers 3,7,1h5. Consider all possi | samples of size two that can be drawn with replacement from this populatig, Find (a) The population mean, (8) The population standard deviation, (©) The mean of the sampling distribution of means, (@) The standard deviation of the sampling distribution of means, | IINTU CK) Nov.2012 (SetNo.34y) | aoe Probability ang gy | 12, Calculate the size of S.D of means for the population 16, 14, 12, 8, 24, 20 by drawing samples of size 2 without replacement. Verify the results. 13. A population consists of six numbers 4, 8, 12, 16, 20, 24. Consider all samples| | of size two that can be drawn with replacement from this population, Find (a) the population mean (5) the population $.D (c) the mean of the sampling distribution of means (d) the standard deviation of the sampling distribution of means. [INTU (K) Dec. 2015 (Set No.2)] 14. A population consists of the four numbers 2, 3, 6, 8. Consider all possible samples of size two that can be drawn with replacement from this population Find (i) the populations mean, (ii) The population standard deviation, (iii) The mean of the sampling distribution of means, [INTU (K) I Sem. June 2015 (Set No. 1)) 15. A population consists of the four numbers 2, 3, 4, 5. Consider all possible samples of size two that can be drawn with replacement from this population. Find (i) the populations mean, (ii) The population standard deviation, (iii) The mean of the sampling distribution of means (NTU (K) If Sem. June 2015 (Set No. 4)] 16. If the population is 3, 7, 9, 11, 15. (@ List all possible sam, from the finite populati distribution of means, (iii) of means. INTU (K) 11 Sem, June 2015 (Set No.3) 1. (@) 0.9799 (4) 0.8361 2.30 3.0.8 4. 0.16 5. (a) Itis divided by 2 () It is multiplied by 4, 6. (i) 0.1806 7.(a) 0.0919 (b) 0.3655 — © scanned with OKEN Scanner iples of size 3 that can be taken without replacement (i) Calculate the mean of each of the sampling Find the standard deviation of sampling distributioranplng_Distbutions 269 P(Z 20-71) = (0.2389) 9.P(510 <¥ < 520) = 0.6553 =20,6= 4.4721 8 0. PUES 715) = 0.0062 11. (i) (a) p=9 (6) p (c) means 3, 5, 7,9, 11, 13, 15 (d) a, =9 (@) o,(o/Vn) a, WaT He = 14, 6 = 6.683, of = 4.32 3. @ual4 (B) 0? = 46.67 (0) wy = 14 (d) 9x iM3 SAMPLING DISTRIBUTION OF THE MEAN (o UNKNOWN) = 18.67, = we + oF In previous problems on a population mean or the difference between two sepulation means it was assumed that the population standard deviation ois known. But for large sample of size (n > 30), even if standard deviation ¢ of population snot known, it does not make any difference. Since we can substitute the sample iD ‘sin the place of and the sample S.D ‘s’, is calculated using the sample mean (xP z by the formula s* z Ss For small sample of size (n < 30), when o is unknown, it can be substituted 1ys provided we make the assumption thatthe sample is taken from normal population F- distribution and % - distribution in Chapter 8 Quiz We will discuss the f - distribution, 1. The totality of the observation is called to) (a) Population (b) Sample (c) Parameter (a) None 1. The statistical constants of the population are called to] (a). Statistic (b) Parameter (c) Sample statistic (d) None 3. ‘The probability distribution of a statistic is called to] (a) Normal distribution (b) Sampling distribution (d) Binomial distribution (d) None 4. The number of possible samples of size n out of N population units without replacement is [ @) MN, (b) Nv 1 © We (a) None © scanned with OKEN ScannerCHAPTER - 6 ESTIMATION | ee 61 INTRODUCTION Consider a random sample of m observations, say ¥5%,,--0%, iS selected from Population whose probability distribution function is /(,0;,03,--»0,) Where 0,,0,,...,0, age the population parameters. Then there will always be an infinite number of functions of samp. values, called statistic, which may be proposed as estimates of one or more of the parameter, The statistic whose distribution concentrates as closely as possible near the true value of the parameter may be regarded as the best estimate, In other words, the problem of estimation cay be written as “determining the functions of sample observations 6,(x,,x),.. +), .x,,) Such that their distribution is concentrated as closely as (i possible near the true value of the parameter”. These estimating functions are also referred to as estimators. 62 DEFINITIONS Quantities appearing in distributions, such as p in the binomial distribution and 4 and in the normal distribution are called parameters, Estimate [JNTU (H) Nov, 2009, (A) 2010S, Nov. 2011 (Set No. 1), May 2012) An estimate is a statement made to find an unknown population parameter. Estimator [JNTU (H) Nov. 2009, (A) 2010S, Nov. 2011 (Set No. 1), May 2012] The procedure or rule to determine an unknown population parameter is called an estimator. For instance, sample mean is an estimator of population mean because sample mean is a method of determining the population mean. Remember that an estimator must be statistic and it must depend only on the sample and not on the parameter to the estimated. So an estimator is a statistic which for all practical purposes, can be used in place of unknowt | parameter of the population. A parameter can have one or two or many estimators. Types of Estimation : Basically, there are two kinds of estimates to determine the statistic of the populatio® parameters namely, (a) Point Estimation and (6) Interval Estimation. Statistical Estimation W. A. Spur and C. P. Bonini defined Statistical Inference as “the process by which 8 draw a conclusion about some measure of a population based on a sample value. The m2" might be a variable, such as the mean, S. D., ete. The purpose of sampling is to estimate $°" characteristics for the population from which the sample is selected”. 272 © scanned with OKEN Scannerjmation on 273 ically, thei tw an ae 2 = 0 oe of problems under statistical inference : (1) Hypothesis testing - to test some hypothesis ab i sample 8 dra. P out parent population from which (2) Estimation - to use the statistics obtained from the sample as esti stimat rameter of the population from which the sample is drawn. if tee An inp problem ot Sita Inference is the estimation of population parameters (je, population mean, population S.D, etc.) from the corresponding sample statistics (i.e. sample mean, sample S.D. etc) point Estimation and Interval Estimation [INTU(K) Nov. 2011, Dec. 2015 (Set No. 4), 11 Sem. June 2015 (Set No. 1) Ifan estimate of the population parameter is given by a single value, then the estimate is called a Point Estimation of the parameter. But if an estimate of a population parameter is given by two different values between which the parameter may be considered to Tie, then the estimate is called an interval estimation of the parameter. Example : If the height of a student is measured as 162 cms, then the measurement gives a point estimation. But if the height is given as (1633.5) ems, then the height lies between 159.5 cms and 166.5 cms and the measurement gives an interval estimation. he ‘The sample mean ¥ is point estimate of population mean 1, sample variance s* is a point estimate of population variance oo. Definition : A point estimate of a parameter @ is a singl computed from a given sample and serves as an approximation of le numerical value, which is the unknown exact value of the parameter. Defin and will be denoted by 6 (read as theta hat). Properties of Estimation : An estimator is not expected to estimate the population parameter without error. ‘An estimator should be close to the true value of unknown parameter. Unbiased and Biased Estimates A statistic is satd to be an umbiased estimator ofthe corresponding parameter if the f the statistic is equal to the corresponding population led a biased estimator of the corresponding ove two cases are called unbiased and biased 1 : A point estimator isa statistic for estimating the population parameter 0 mean of the sampling distribution o parameter, Otherwise the statistic is ca parameter, The values of statistics in the al estimates respectively: ¢ the corresponding parameter and £(¢)=O, then ¢ is an If t bea statistic and @ b se { is a biased estimator of @ and the bias is E()—0. unbiased estimator of 0. Otherwi sample mean ¥ is an unbiased estimator of population mean For example, [since E@) =H]: [INTU(A) Nov. 2011 (Set No. 2)] Proof : Let ttpveonty bea Fandom sample drawn from a given population with 2 mean t and variance . Then © scanned with OKEN ScannerProbability and sia, y 274 gg AP E(x) + E(x) +. F EQ) n 1 Sout t Hd fo BG) =H) Hence the sample mean ¥ is an unbiased estimator of the population mean | Unbiased Estimator : Let 6 be an estimator of 9. The statistic @ is said to be unbiased estimator, or its value an unbiased estimate, if, and only if the mean or expecteq Value of 6 is equal to @. This is equivalent to say that the mean of the probability distributing of 6 (orthe mean of the sampling distribution of 6) is equal to @. An estimator possessing i, property is said to be unbiased. Definition. Unbiased estimator : A statistic or point estimator 0 is said to be an unbiased estimator of the parameter 6 if E (6) In other words, if E (statistic) estimator of the parameter. Example 1: Show that $? is an unbiased estimator of the parameter o?. Solution: Let us write Do-w = Y fei in =z a 2 i parameter, then statistic is said to be an unbiased [@;-0)-@-wP Hw) > Gy -w+n@ ny? -yy-2 % fal (4-H)? =n =n)? ve (1) 3 (x, - 3 Now ES) = E x a a x Bown 6-0 |, ing a However, o%, = ES) © scanned with OKEN Scanneryr estimation 275 Although $? is an unbiased estimator of «?, S, on the other hand, isa biased estimator of with the bias becoming insignificant for large samples. This example illustrates why we givide bY (n-1) rather than 7 when the variance is estimated. Example 2: Prove that for a random sample of size n, X,,Xay.~-»%» taken from an 2_1¢ oe . . population s =U -X" is not unbiased estimator of the parameter o” but infinite ial EEO —X)° is unbiased. [NTU (HH) Noy. 2010, (K) Dec. 2015 (Set No.2,3)] alist Solution: Let 1 be the population mean. Then B(x) =m and var(x,) =B((x,—1)']=0? for 1=1,2,. Sey -F? ~@ {fs be the sample S. D., then s* 2p 28 a nth? +H = 2 = mn «-W 2 Etsy -w [sew ; oe =e[ 2M -c-»"] =f =v OD _ var) n Thus E(?)#o° MOT Wi aE © scanned with OKEN Scannerov. 276 Probabilly and Stags cy Hence, s? is a biased estimator of o? Let us consider $? de-a 8 [by (1)] n- n Then E(S*) = x & a 1 - 2 Hence S$? =" )°(x, -z)’ is an unbiased estimator of 0”. Variance of a Point Estimator : If 6, and 6, are two unbiased estimators of the sang population parameter @, we would choose the estimator whose sampling distribution has the smaller variance, Hence, if ©}, <5}, , we say that 6, is a more efficient estimator of @ than Most Efficient Estimator : If we consider all possible unbiased estimators of some parameter @, the one with the smallest variance is called the most efficient estimator of 9. [NTU (A) Nov. 2011 (SetNo.2) In Figure, we illustrates the sampling distribution of three different estimators 6, and 6 and 6, all estimating @. Itis clear that only 6, and 6, are unbiased, since their distributions are centered at @. The estimator 6, has a smaller variance than 6, and is therefore more efficient Hence our choice for an estimator of @ , among the three considered would be 6,- 6 ° Fig. Sampling Distribution of Different Estimation of 9 Properties of Estimators : [INTUCK) Dec. 2013 (Set No.3)! A good estimator is one which is as close to the true value of the parameter as possible The important properties of @ good estimator are : (@ Consisteney (i) Unbiasedness (ii) Efficiency and (iv) Sufficiency (@ Anestimator 8, of a parameter @ is consistent if it converges to 9, a8 17° (Z)A statistic 6 is said to be an unbiased estimate of 0 if E (6,)=0 forall 0- © scanned with OKEN Scanneration esti Tn (iii) Astatistic 0, is said to be a more efficient unbiased estimator of the parameter 0 than the statistic 6, if (a) 6, and 6, are both unbiased estimators of @ ® V@)< VEG) (iv) An estimator is said to be sufficient for a parameter, ifit contains all the information jn the sample regarding the parameter. 6.3 INTERVAL ESTIMATION [UNTU (A) Nov. 2011 (Set No. 3,4)] Point estimates rarely coincide with quantities they are intended to estimate. So instead ofpoint estimation where the quantity to be estimated is replaced by a single value a better way of estimation is interval estimation, which determines an interval in which the parameter lies. Interval Estimate : Even the most efficient unbiased estimator cannot estimate the population parameter exactly. It is true that our accuracy increases with large samples. But there is still no reason why we should expect a point estimate from a given sample t0 be exactly equal to the population parameter, it is supposed to estimate. Therefore in many situations it is preferable to determine an interval within which we would expect to find the value ofthe parameter. Such an interval is called an interval estimate ‘Thus an interval estimate is an interval (confidence interval) obtained from a sample. Interval Estimation : An interval estimate of a population parameter 6 isan interval of the form 6, < @ < 6,,, where 6, and 6, depend on the value of the statistic 6 for a particular ampling distribution of 8. sample and also on the si Il generally yield different values of 6 and, therefore, different Since different samples wil values , and 6, . These end points ofthe interval are values of corresponding random variables and 8, 4, and 6,. From the sampling distribution of 6 we shall be able to determine 6, and 6, such that the P(6, , ~ p2)) Note. If for a statistic, P(-3<2<3) i.e., confidence limits cover 99.73% area under the standard normal curve, then we say that the confidence limits . for statistic are almost sure limits without mentioning the degree of confidence. te In calculating S.E. if population parameters are unknown, corresponding sang, statistics are used to find an approximate value of S. E. For example, if the poputaj, S.D., is not given, the S.E. of ¥ is calculated by using sample S.D. s in place of ¢ 6.5 DETERMINATION OF PROPER SAMPLE SIZE Determination of proper sample size is important for testing of hypothesis in busines problems. The size of sample should neither be too small nor too large. If the size oft sample is too small, then it may not give a valid conclusion. On the other hand, if the size of the sample is too large, then there may be loss of time and money without getting the required results. 1. Sample size for estimating Population Mean Let ¥ be the mean of a random sample drawn from a population having mean 4 and $.D. c. Let the sampling distribution of the sample mean ¥ be approximately normal distribution with mean #1 and S. D. o. If E be the permissible sampling error, tha E=¥-p. ‘The confidence interval for the population mean #1 is ¥+2(S.E. of ¥) =74E where z= confidence coefficient or z~ value (which is 1.96 at 5% level of significa) and E=z(S.E. of ¥) = sn being the sample size and ¢ = population S.D. 2. Sample size for Estimating Population Proportion Let p be the sample proportion and P be the population proportion. If £ ; permissible sampling error, then E'= p—P and the confidence interval for the PoP" [PO pve 1 2 proportion Pis p+z(S.E. of p)=p+E where E= (S.E. of p) = 2 sample size, or, E7=2 72 4, ,2PO ” where Q=1-P. >| © scanned with OKEN Scannerao u nf ty gstimation Bat Proposition : If x|,x2,..,., be a random sample from an infinite population with ; 2 . Ie variance o* and sample mean X, then —)°(x,—¥)? is a biased estimate of o?, but a o xy ;. . DG; -%)* is an unbiased estimate of 0”. 3.Maximum Error of Estimate E for Large Samples: [JNTU (A) Dec. 2009 (Set No. 2)] Since the sample mean estimate very rarely equals to the mean of population #, a point estimate is generally accompanied with a statement of error which gives difference between estimate and the quantity to be estimated, the estimator. Thus error is | ¥ — H|. For large 7, the random variable *— is normal variate approximately. ofvn Then P (- za <2 < zap) =1— @ where z= 2—# GC tap 2) a elm z= Hence P| -24/2 < 30), one can assert with probability (1 — a) that the Pid Ty |¥— | will not exceed ran(F): in Sample size: When a, E, care known, the sample size 7 is given by a = [225 o> (3) | When ois unknown : In this case, o is replaced by s, the standard deviation of sampk | to determine E. ‘Thus the maximum error estimate E=fy2 Fz with (Ia) probability. In 4. Maximum Error of Estimate E for Small Samples : When 2 <30, small sample, we use S, the S.D of sample to determine E, When o iss! known, f can be used to construct a confidence interval as 1. The procedure is the same as that with known a except that o is replaced by Sandie standard normal distribution is replaced by the /-distribution. P(laj2 <1 < faj2)=1- & Fig. Plan <1< fan) =1- a ~ cove which we find Where tas is the ¢-value within (7 — 1) degrees of freedom ab 1 to He of a/2. Because of symmetry, an equal area of a/2 will fa of —ta2. © scanned with OKEN Scanner