Professional Documents
Culture Documents
oJSTRIBUllONS
,&r.4pL1NG . f rence about a certain phenomenon, sampling is a well accepted tool. Entire population cannot be
I :,,: d aw in e · · 1· · h f I·
IJ oilier 10 reral reasons. In such a situation samp mg 1s t e onlt alterna~ive. A properly drawn sample is mu~h use u. in
due 10 seV ions. Here, we draw a sample from probability d1stnbution rather than a group of obJects. Using
,.d I ,oneIus
;-': reliabe_ sample is drawn .
.,,n9 . ,h que • "b t·
:•· o0ntev· 01 continuous D1str1 u 10n
,,J a ..
'.,.,,,nSI"· pie frOll'I a
le from a continuous pro bab'I'
11ty d'l~tn'but1on
. f (x, 8) is nothing but the values of independent and 'd . II
I entica Y
~ -,andoiTl sannP 'ables with the common probability density function f (x, 8).
,. donn vari . . . . .
,s11bllted ran . If Xi, x , ....., Xn are independent and 1dent1cally distributed random variables, with p.d .f. f (x, 8), then we say
2
· Dlfinld0" • pie from the population with p.d.f. f (x, 8).
fornn asann
!,t~ey
io~: •ng inference, we use the numerical values of X1, x2, ...., Xn.
l for draw1 .
. . t pd f of X1, X2, ... Xn 1s,
TheJOIO · · ·
n
ll f (Xi)
i=1
~nc AND PARAMETER - -··. - · - · I
l.lUsing;random sample Xi, X2, ..., Xn we draw conclusion about the unknown probability distribution. However probability
, ...,, - can be studied if the parameter 8 is known. In other words study of probability distribution reduces to the study of
ai5tnw1100
rarneter a. we use sampled observations for this purpose. There are various ways of summarizing the sampled observations.
:esummarized quantity is called as statistic. We define it precisely as follows.
Definition: If X1, X2, ..., Xn is a random sample from a probability distribution f (x, 0), then T = T (x1, x2, ... , xnl a function of
11111
plevalues which does not involve unknown parameter 0 is called as a statistic (or estimator).
Some typical statistics are given below :
2Xi
Ii Sample mean : n
T = X is a statistic
[ii) Sample variance :
T = T (xi, x2, ... xn)
1
- n_
1
2 (Xi - -x )2 is a statistic.
~SQUARE DISTRIBUTION
Introduction:
,ppl~ei~~i-square (pronounced as Ki, sky without 's') distribution is one of the important distributions in Statistics. It is mainly
11t
ing of hypothesis for testing the independence of attributes, testing the goodness of fit of a model etc
Th •
echi-square . . 2
variable 1s denoted by Xn . Hence n is the parameter of the distribution, also, called as the 'degrees of
1
~om· (d.f.). The 2 . . .
Xn variate 1s defined as sum of squares of n independent standard normal [N (0, 1)] variables.
"J. • • • ,it. • . .. , "fl '-''- "IIIV <;;/'-'C ll UC ll l l't \ U , .J.J Vd ll d U IC .> , lll '-"
n
Y L 2
Xi follows chi-squa re distribution with n degrees of freedom (d.f.).
i= l
2
Notation : Y --> Xn . (·: in positive integer)
In this text we will consider null hypothesis to be hypothesis of equality and alternative hypothesis to be two sided. Choice of
one sided hypothesis as null hypothesis is beyond the scope of the book.
~-
ut VJ. . . region (W) and acceptance region (WC) as shown below Fig. 7.2S
ets : cnt1ca 1 ·
e, proportion
stJbS Of HYPOTHESIS
or tw · For 7JJArule wh1c. h leads .to the decision of acceptance of H0 or re;ect·
, ion °f Ho on the basis
• of observations
. in a random sample ·1s
o states etc.
t of hypothesis.
Statistical inference .15 th.at_branch of stat,.5t .'cs which
. concerned with using probability concept to deal with uncertainly in
.. n making field of statistical inference has a fruitful development since the letter half of the 19th century. It refers to the
:al distribution.
-e .
deC'''0, of selecting and using a sample statistic to draw inference about a population parameter based on a sub-set of it the
p,ocelsdrawn from the population. Statistical inference treats two different classes of problems
') Hypothesis testing and (ii) Estimation.
~ypothesis testin~ . begins with us assumption called as hypothesis, which is made by the population parameter. A
! ,s rejected, the
thesis is a suppos1t1on made as basis for reasoning.
~n article 7.14 to 7._21 _we discussed the various terms like statistical hypothesis, null hypothesis and alternative hypothesis,
Jil hypoth~sls is critical region, level of s1gnif1cance etc. . .
Now we will see the method of testing population mean (µ) equal to specified value (µ 0).
Testing Population Mean (µ) Equal to Specified Value (llo) : Test statistic is
JO. Ho: µ1 = µ2,
! difference will
:ided in testing
(- 2J -N
k k oi
,,_2 = I = I
i =1 i = I ei
has ,,_2 distribution with (k - p - 1) degrees of freedom. Degree of freedom is a parameter of the x2 distribution. In th·
iscas,.~
critical region at level of significance a is
2 2
Xk - p - 1 'X.k - p - I; a
where, 'X.~ _ P _ I ; " is the table value corresponding to degrees of freedom k - P - 1 and level of significance (l.o.s) Q We~/
2
. . I va Iue.
'X.k _ P _ 1, a as a cntIca
It is shown by the shaded region in Fig. 7.26.
Normally, the values of a are taken as 0.05 or 0.01.
Thus, we reject Ho at 1.o.s. a if,
2 2
Xk- p -1 Xk - p -1; a
y
Density curve
otl
oo X
(0, 0)
- - - - -(C.R.}--
Rejection Region for H0
2
Xk-p-1 : a
Fig. 7.26
Note: ·ii i
1. We can apply this test if expected frequencies are greater than or equal to 5 (i.e. ei 5) and total of cell frequen~
sufficiently large (greater than 50). f
. I 'th its obsen .
2. W hen expecte d f requency o f a cIass Is less than 5, the class is merged into neighbouring class a ongwi 10
th
and expected frequencies until total of expected frequencies becomes 5. This procedure is called ·pooling e'
In this case, k is the number of class frequencies after pooling. . ect~
3. It is obvious that if any parameters are not estimated while fitting a probability distribution or obtaining exP
frequencies, the value of p is zero. C\lnt
4. This test _is not applicable for testing goodness of fitting of straight line or curves such as second degree
exponential curve etc.
Remark : eeted ini
2
~~te's Co~ectlon : If in_a 2 x 2 conti~gency table, any cell frequency is less than 5 then the test statistic X is
spec1f1c way. This correction 1s due to Yates and hence is known as Yate's correction. It is beyond the scope of thiS b
74
r,1ATICS-lll"(~CIV1::,::::Ll_ _ _ _ _ _~;~;;~
(.:..:
'.'.'·:::.
_:!.
S..;); : - = . ~ - - - -_!'.P~R~0!_BA~B~l~UTY~!_
A~N~D~P".!R~O~B~A~B~IUTY
~~D~IS~ T~Rl~B~U:._:
Tl::O
:.:N
~S
Gr,1Aft1E - '-
(jlllll~ l1LLUSTRATIONsj
11
. { ed bank utilizes four teller windows to render fast service to the customers. On o particular day, 800
: ;,. notiona : They were given service ot the different windows os folio WS :
£;• J
were observe . Wlndow Number Expected Numbe; 4 customers
~~""ers 1 150
2 250
3 170
4 230
We reject H0 at 5 % l.o.s.
Conclusion: The customers in the nationalized bank may not be uniformly distributed over different windows.
Ex, 2: One hundred samples were drawn from a production process each a~er 5 hours. The number of defectives in these
samples were noted. A Poisson distribution by estimating the parameter m was fitted to these doto. The results obtained are as
follows :
3 2 1.26
4 1 0.16
5 and above 0 0.02
Test the go0 d
Sol . ness of fit of Poisson distribution in above situation. [Use 5 % level of significance]
• We want to test
Ho : Fitting of Poisson distribution is good {proper) against
H1: Fitting 0f p · . . . .
01sson d1stnbut1on 1s not proper.
TICS_ ID (CIVIL) PROBABILITY AND PROBABILITY DISTRIBUTIONS
,t!:-11~
. ed bank utilizes four teller windows to render fast service to the customers. On a particular day, 800
J : r, natiana a · service· a t th e d 1,,erent
1 '" ·
d They were given windows as follows .
i,, were observe . Window Number Expected N umbEf ,1{ customers
~ ers
1 150
2 250
3 170
4 230
ustamers are uniformly distributed over the windows.
hetherthe c . . . .
rest W we want to test Ho : Customers are uniformly d1stnbuted over the windows. i.e. Ho : customers on all windows are
so • .
L,Here They are not equaI on a II win
. d ows.
ainstH1- .
equal ag h xpected frequencies are :
under H~ t e e
Window Number Expected Number of Customers
(ej)
1 200
2 200
3 200
4 200
Thetest statistic is
2 (-50)
2
illt (-30)
2
fill:.
Xk- p -1 200 + 200 + 200 + 200
i= l
We reject Ho at 5 % 1.o.s.
Conclusion: The customers in the nationalized bank may not be uniformly distributed over different windows.
Ex. 2: One hundred samples were drawn from a production process each after 5 hours. The number of defectives in these
samples were noted. A Poisson distribution by estimating the parameter m was fitted to these data. The results obtained are as
follows :
Number of Number of Samples (observed) Expected Number of Samples
Defectives
0 63 60.65
1 28 30.33
2 6 7.58
3 2 1.26
4 1 0.16
5 and above 0 0.02
th
:t e goodness of fit of Poisson distribution in above situation. [Use 5 % level of significance]
I. : We want to test
H.0 F'tf
1
· ing of Poisson distribution is good (proper) against
H1 : Fittin
9°fp · . . . .
0Isson d1stnbut1on Is not proper.
r
ENGINEERING MATHEMATICS - m (CML) ( 7A 6) PROBAIIIUTY AHO P R o ~
Here we pool expected frequencies until their sum becomes 2: 5 and also pool corresponding observed f ~-
frequencies can be written as : reque•-~
Flcies. T~
Observed Frequ1111cles Expected frequencies ltt~
(ol) (el)
63 60.65
28 30.33
9 9.02
We use the test statistic
2
Xk- p-1
i=l
i (iJ-N
Here number of parameters estimated = p = 1, N = 100, k = 3.
2
X1 100.27009 - 100
2
X1 0.27009 [Calculated value]
2
X1 0.27009 < X~; o.o5 = 3.841 [Critical or Table value]
Hence we accept Ho at 5 % l.o.s.
Conclusion : Fitting of Poisson distribution may be good to the given data.
Ex. 3: Among 64 of{springs of a certain cross between guinea pigs 34 were red, 10 were black and 20 were white A;-;-
a genetic model these numbers should be in the ratio 9 : 3 : 4. · ' 0' ~g10
Are the dato consistent with the model at 5 % level ?
Sol. : Here Ho : The offsprings red, black and white in the colour are in the ratio 9 : 3 : 4.
In this pro blem, N = 64. Hence observed and exoecte d f reauencies are as f ollows :
Observed Frequencies (oj) 34 10 20
Expected Frequencies (el) 9 3 4
X 64 = 36
16 16 X 64 = 12 16 X 54 = 16
To test H0, the test stat1st1c is
k
2 , (oi-ei) 2
Xk-p-1 L ei
Here p =0 and k = 3.
i=l
2
X2 1.444444 (Calculated value)
2
X2 o.o5 5.991 (Critical o r Table value)
2
X2 1.444444 < X~; o.05 = 5.99
We accept Ho at 5 % l.o.s.
Conclusion : The data are consistent with the genetic model that the offsprings red, black and white in colour are in th e ratio
9: 3 : 4.
Ex. 4 : The table below gives n umber of books issued fro m a certain libran on the various doys ofa week.
Days No. of Books Issued (01- e,P
Mon. 120 0
Wed. 130 100
Thr. 110 100
Fri. 115 25
Sat. 135 225
Sun. 110 100
Test at 5 % l.o.s. whether issuing the book is day dependent.
·••TICS-Ill (CIVIL) (7A7)
PROBABILITY AND PROBABIUTY DISTRIBUTIONS
""'flll,..,.
-~ . of the book is not dependent on the day of the week.
~ t,e issu1n9 2 z
,.1. : 2 = X&-0-1 = Xs
,.. X1c-p-1
2
L (oi-eil2
Xs = ei
2 550
Xs = 120 = 4.5833
2
Xs. o.os = 11.07
2 2
X5 < X5, o.o5
Accept Ho-
• · e issuing of book is independent of day.
condusIon , 1• ·
fl. : experiment on pea breeding, the following frequencies of seeds were obtained
5 1n Total
Wrinkled and Green Round and Yellow Wrinkled and Yellow
Roundan_
d,
gl'ffll .
e,pinment.
Sol: from the given data the corresponding frequencies are, i.e. expected frequencies are
2 2 1
Expected 8
l3 524 = 81 l3 X 524 = 81 l3 X 524 = 40
l3 X 524 = 323 X
Frequencies (el)
2 2 1 (Oi-eil2
Xk-p-1 X3 ei
2
2 (222 - 323) 2 (120 - 81) 2 (32 - 81) 2 (150 - 40)
X3 323 + 81 + 81 + 40
&. &:A set of five similar coins is tossed 210 times and the result is
1 2 3 4 5
No.of Heads 0
5 20 60 100 23
Frequency 2
th (DK.2009)
Test ehyPorhesis that the data follow a binomial distribution.
Sol : Here k - p - 1 = 5.
1
p Probability of getting a head = 2
1
q Probability of getting a tail = 2.
(7.48) PROBABIUTY AND
ENGINEERING MATHEMATICS- m (CML) PRoaAalUT-r
. • f tt · o 1 2 3 4 5 heads are the successive terms of th b. D1Sflllaui,.
Hence the theoret,cal frequene1es o ge ing , , ' ' ' e inornial '-.."ii
q) s 210,
2 2 3 5 4 5
= 210 [p5 + 5 p4 q + 10 p 3 q + 10 p q + pq + q ] '·
l 5 10 10 2- + ..1.J
= 210 [ 32 + 32 + 32 + 32 + 32 32
7 + 33 + 66 + 66 + 33 + 7
The theoretical frequencies are 7, 33, 66, 66, 33, 7.
2 (2 - 7)2 (5 - 33) 2 (20 - 66) 2 (60 - 66) 2 (100 - 33)2 Q.3 _ ?i
Hence, X5 7 + 33 + 66 + 66 + 33 + 7
2
X5 3.57143 + 23.7576 + 32.06061 + 0.5455 + 136.0303 + 36.5714
2
Xs 232.53684
n
20
6 !}
18
7
l
4
0.1429
0.2222
28 25 9 0.36
42 40 4 0.1
22 25 9 0.36
15 18 9 0.5
;}
2
7
n 7 0 0
We take table value of x:. Because, we have total 9 classes, 2 classes are pulled, so after pulling number of classes is 7.
Now, we have,
2 2
X6 < X6. o.o5
Accept H0 .
_ _C::.o::.nc.:.c::.l.::u.::.s:..:io:..;.n.:...c...:T.:_h..;.e:......:fi.:_tc.:is__,ge.co:..:o:..:d::.·_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
___
d
,,cs- Ill (CJVILl (7 A 9 ) PROBABIIJTY AND PROBAlllUTY DISTIUJIUTIONS
ft!~ . .
'1'.G IP' r particular spare port 111 a factory was found to vary from day to day In O sample study the fo llowing
# , ,·,nede,r,andu,ed.fo O
1 0
~n ,.as obt c,oys~- - - -+-_
M_o_n._ + _Tu_e_s_.-1:--..:.W
:.:e:.:d:_._Jf-.:_:
Th~u::rs=. _f----.'.Fr
..:'..'i._-4-_.::S:.
.: at::..--,
/' .,,,,u .,o.,na~:=:.:n_d_e d_ __,__l _l_24_ _.__l_l _2 _5 _L.__l_l_l _O_J.__:l:.:l:.:2_:0_.1.__l::l:.:2:.:6:.___L~l_l _l5_ __,
No-ofr• -
. t the number of parts demanded does not depend on the day of the week.
0
theSLS thO
rJ,ehYP
rill ,si·
L: 10 1 · rtS demanded does not depend on t he day of the week.
So urnber of pa
;;:: fhe n ber of parts demanded depend on the day of the week.
. n,e nurn . .
vsH, · rts demanded dunng six days = 6720.
,nt,erof pa
· of parts to be demanded each day of the week = - 6- 1120.
E#'1ed no.
:--
oays 0; ei
I
(o,-e,>1 (0;-e.)2/ei \I
,--
Monday 1124 1120 16 0.01429
Now,