Professional Documents
Culture Documents
Keywords- Test Statistic, Z-test, Student’s t-test, F test (like ANOVA), Chi-square test, Research Methodology
⁄ )⁄ ⁄ ) Chi- As ,
Proof: Independent squared ⁄
̅ ̅ ) As ,
⁄√ ⁄ Normal
( )
⁄√ ∑ ̅) ⁄
Note: Let ), ,
̅ )
∑ ̅) ⁄ III. TEST STATISTIC APPROACH OF
HYPOTHESIS TESTING
̅ )
̅ ) ⁄ Hypotheses are predictions about the relationship among two
∑ ̅) ⁄ ∑ ̅) or more variables or groups based on a theory or previous
⁄ ) research [1]. Hence, hypothesis testing is the art of testing if
variation between two sampling distributions can be
explained through random chance or not. The three
approaches of hypothesis testing are test statistic, p-value and
Proof: Consider all of the possible samples of size n from a Lemma 2. If and are
population with expected value ) and variance ). independent normally distributed random variables with
If a sample are chosen, each comes from ) and ) , then ̅ ̅
the same population so each has the same expected values,
) and variance ). ).
Proof: By Lemma 1. ̅ )and ̅ ). Then
̅) ( ∑ ) (∑ ) (∑ ))
̅ ̅ ̅ )̅ ) ).
)) ) The appropriate test statistic for this test is given by:
̅ ̅)
) (6) Theorem 3. If and are normally
√
distributed random variables, but and are dependent,
then the standard random variable is:
̅
where ∑ ̅ ) and ∑ ( ̅) ) (10)
⁄√
If we do not know the variance of both distributions, then we
may use: where is a pairwise difference (is the sample
̅ ̅) )
mean of the sum of differences between the first results and
√
(7) the second results of participants in the study) also given by
̅ ∑ , is the mean value of pair wise difference
provided n and m are large enough to invoke the Central under the null hypothesis, and is the population standard
Limit Theorem, where is the sample size of the one group deviation of the pairwise difference. The basic reason is that
(X) and is the sample size of the other group (Y), ̅ is the and are normal random variable, so is
sample mean of one group and calculated as: also normal random variable, thus are thought
to be a sequence of random variable.
∑
̅ , and the sample mean of the other group is
B. Student’s t 𝓽-test
denoted by ̅ and can be estimated as: Unfortunately, -test requires either the population is
∑ normally distributed with a known variance, or the sample
̅ size is large. However, when the sample size is small
), the value for the area under the normal curve is
Two sample Z-test (when Variances are Equal) for Mean not accurate. Thus, instead of a z value, the t values are more
It can be applied to investigate the significance of the accurate for small sample sizes since the value depend upon
difference between the mean of two populations when the the number of cases in the sample, i.e., as the sample size
population variances are known and equal, but both change; the value will change [5]. However, the -test
distributions are normal. The null and alternatives hypotheses needs an assumption that the sample value should be drawn
(two-tail test) are given by and randomly and distributed normally (since the data is
, respectively. Then test statistic is: continuous). Therefore, the -test is a useful technique for
̅ ̅) comparing the mean value of a group against some
) (8)
√ hypothesized mean (one-sample) or of two separate sets of
numbers against each other (two-sample) in case of small
where standard error of difference is √ samples ( ) taken from a normal distribution of
unknown variance.
If we don't know the variance of both distributions, then we
may use: A t-test defined as a statistical test used for testing of
̅ ̅) )
(9) hypothesis in which the test statistic follows a Student’s
√
t-distribution under the assumption that the null hypothesis is
provided and are large enough to invoke the Central true [6].
Limit Theorem.
One-Sample t-test
Paired Sample z Test Banda Gerald [4] suggested that a one-sample -test can be
According to Banda Gerald [4], the words like dependent, used to compares the mean of a sample to a predefined value.
repeated, before and after, matched pairs, paired and so on are Thus, we use one sample t-test to investigate the significance
hints for dependent samples. Paired samples also called of the difference between an assumed population mean and
dependent samples used for testing the population mean of a sample mean. The population characteristics are known
paired differences. A paired -test helps to determine from theory or are calculated from the population. Thus, a
whether the mean difference between randomly drawn researcher can use one-sample t-test to compare the mean of a
samples of paired observations is significant. Statistically, it sample with a hypothesized population mean to see if the
is equivalent to a one-sample -test on the differences sample is significantly different. The null and alternative
between the paired observations. The null and alternative hypotheses (two-tail test) are given; and
hypotheses (two-tail test) are given by and , respectively. Then the t-test statistic is:
̅
, respectively. ⁄√ ) (11)
Equal)
A two-sample t-test is used to compare two sample means Theorem 4. If and are normally
when the independent variable is nominal level data while the and independent distributed random variables with
dependent variable is interval/ratio level data [7]. Thus, ) and ), then the test statistic is
independent samples -test is used to compare two groups given as:
whose means are not dependent on one another. Two samples
̅ ̅)
are independent if the sample values selected from one (13)
population are not related with the sample values selected √
from the other population. Thus, an independent sample t-test ∑ ̅) ∑ ̅)
) )
helps to investigate the significance of the difference between where √ √
the means of two populations when the population variance is has a -distribution with degree of freedom.
unknown but equal.
Proof:
Let assume a group of independent normal samples given by ̅ ̅)
and distributed as
) and ), respectively. Then, the √ ) )
√
null and alternative hypotheses (two-tail test) are given by
and , respectively. If ̅ ̅) ) )
sample sizes and are not large enough to invoke the ⁄√( )
Central Limit Theorem, then the test statistic is: √
̅ ̅)
) (12) ̅ ̅)
√ Let , by lemma 4. ) and if
√
where refers to sample size from the first group (X), is
) )
sample size from the second group (Y), ̅ is the sample mean ( ).
of the first group (X), ̅ is the sample mean of the second ) )
Then since and .
group (Y), and is the pooled standard deviation for both
groups which is calculated as follows: Therefore, )
√
two-tail case, our critical region of size will be all values of D. The F-test Statistic
such that either ⁄ or ⁄ . F-tests are used when comparing statistical models that have
been fit to a data set to determine models that best fit the
Chi-Square Test of Independence population from the data being sampled [15]. F distribution is
a joint distribution of two independent variables with each
The chi-square test of independence is a non-parametric
having a chi-square distribution [16]. F Distribution needs
statistical test used for deciding whether two categorical both parent populations to be normal, have the same variance
(nominal) variables are associated or independent [13]. Let and independent sample. F-test is applicable for testing
the two variables in the cross classification are X and Y, then equality of variances and testing equality of several means.
the null against
alternative . The
chi-square statistic used to conduct this test is the same as in Testing for Equality of Variances
the goodness of fit test given by: The chi-square test used for testing a single population
( )
variance, while -test for testing the equality of variances of
∑ ∑ ) ) (19) two independent populations (comparing two variances).
Thus, it helps to make inferences about a population variance
where is the test statistic that asymptotically approaches a or the sampling distribution of the sample variance from a
chi-square distribution, is the observed frequency of the normal population given the null against the
alternativehypothesis . Suppose two
ith row and jth column, representing the numbers of
independent random samples drawn from each population
respondents that take on each of the various combinations of
(normally and identically distributed population) as given by
values for the two variables.
) ) ) and ). Moreover, let and
represent the sample variances of two different populations. If
both populations are normal and the population variances
where is the expected (theoretical) frequency of the ith and are equal, but unknown, then the sampling
row and jth column, is the number of rows in the distribution of is called an -distribution.
contingency table, and is the number of columns in the
contingency table. The first samples are normally and
If )
) ) , then the null hypothesis is rejected independent distributed random variables,
and we conclude the existence of relationship between the ∑ ̅)
variables. If ) ) , the null hypothesis cannot
with d.f.
be rejected and conclude that there is insufficient evidence of The second samples are normally and
a relationship between the variables. ) ∑ ( ̅)
independent random variables
A one-way analysis of variance (ANOVA) is given by: hypothesis tests to the right data is the main role of statistical
tools in empirical research of every science. Specifically, the
where th th
is the j observation for the i income level, is researcher apply -test, which is appropriate to test the
a constant term, is the effect on the ith income level, existence of population mean difference in the case of large
) . sample size, while t-test is used for testing the mean value
The null hypothesis given by
with independent and dependent small sample size.
against .
Total variation of the data is partitioned as variation within Moreover, -test is used for test of several mean and
levels (groups) and between levels (groups). Mathematically, variance, while Chi-square test is for testing single variance,
̅ ∑ ∑ and portion of total variance, goodness of fit and independence. In general, the researcher
∑ ∑ ( ̅) try to address on the problems of identify an appropriate test
statistic for data analysis, which is the main problems of
∑ ̅ ̅) ∑ ∑( ̅)
researchers who apply a statistical test in research
̅) methodology.
where ∑ ∑ ( is sum of square
variation about the grand mean across all n observations,
∑ ̅ ̅) is sum of squared deviation for ACKNOWLEDGEMENT
each group mean about the grand mean, ∑ ∑ (
̅) is sum of squared deviation for all observation The author thanks the anonymous referees for their valuable
within each group from that group mean, summed across all comments.
groups. Then the mean sum of square between or
∑ ̅ ̅)
denotes the sum of squares that represents the REFERENCES
variation among the different samples, while within mean
∑ ∑ ( ̅) [1] David, J. Pittenger, "Hypothesis Testing as a Moral Choice",
sum square or which refers to the Ethics & Behavior, Vol. 11, Issue. 2, PP.151-162, 2001, DOI:
sum of squares that represents the variation within samples 10.1207/S15327019EB1102_3.
that is due to chance. Then the test statistic is the ratio of [2] Vinay Pandit, "A Study on Statistical Z Test to Analyses
group and error mean squares: Behavioral Finance Using Psychological Theories", Journal of
Research in Applied Mathematics, Vol. 2, Issue. 1, PP. 01-07,
(22) 2015, ISSN(Online), 2394-0743 ISSN (Print): 2394-0735.
[3] Joginder Kaur, "Techniques Used in Hypothesis Testing in
Table 2: One-Way ANOVA Table Research Methodology A Review", International Journal of
Source of Sum of Science and Research (IJSR), ISSN (Online): 2319-7064 Index
DF Mean squares Copernicus Value: 6.14 | Impact Factor: 4.438, 2013.
Variation square
[4] Banda Gerald, "A Brief Review of Independent, Dependent and
Between
SSB k-1 MSSB= One Sample t-test", International Journal of Applied
(between group)
Mathematics and Theoretical Physics, Vol. 4, Issue. 2, PP. 50-54,
Within (error) SSW n-k MSSW= 2018, doi: 10.11648/j.ijamtp.20180402.13.
[5] McDonald, J. H, "Handbook of Biological Statistics", 3rd ed.,
Total TSS n-1
Sparky House Publishing, Baltimore, Maryland, 2008.
[6] J. P. Verma, "Data Analysis in Management with SPSS
Reject if the calculated is greater than the tabulate Software", Springer, India, DOI 10.1007/978-81-322-0786-3_7,
value ( ), meaning variability between groups is large 2013.
compared to variation within groups, while fail to reject [7] Philip, E. Crewson, "Applied Statistics", 1st Ed., United States
if the calculated is less than the tabulate value ( ), Department of Justice, 2014.
[8] Kousar, J. B. and Azeez Ahmed, "The Importance of Statistical
meaning the variability between groups is negligible
Tools in Research Work", International Journal of Scientific and
compared to variation within groups. Innovative Mathematical Research (IJSIMR), Vol. 3, Issue. 12,
PP. 50-58, ISSN 2347-307X (Print) & ISSN 2347-3142
IV. CONCLUSIONS (Online), 2015.
In this paper, the derivation and choice of appropriate a test [9] Tae Kyun Kim, "T test as a parametric statistic". Korean
Society of Anesthesiologists, ISSN 2005-6419, eISSN
statistic were reviewed since application of right statistical 2005-7563, 2015.
[10] Pallant, J., "SPSS Survival Manual", A Step by Step to Data AUTHOR PROFILE
Analysis Using SPSS for Windows (Version 15), Sydney,
Allen .and Unwin, 2007. Mr. Teshome Hailemeskel Abebe working as lecturer in the
[11] Pearson, K., "On the criterion that a given system of deviations
deprtment of economics, Ambo University, Ambo, Ethiopia.
from the probable in the case of a correlated system of variables
that arisen from random sampling", Philos, Vol. 50, pp.157-175, He has 7 years teaching experience. He has a B.A. degree in
1900,.Reprinted in K. Pearson (1956), pp. 339-357.
economics from Debre Markos University and M.Sc. degree
[12] Sureiman Onchiri, "Conceptual model on application of
chi-square test in education and social sciences", Academic in Satatistics (Econometrics) from Haramaya University. His
Journals, Educational Research and Reviews, Vol. 8, Issue. 15,
area of interest is econometrics, especially financial and time
pp. 1231-1241, DOI: 10.5897/ERR11.0305; ISSN 1990-3839,
2013. series econometrics.
[13] Alena Košťálová, “Reliability and Statistics in Transportation
and Communication” In Proceedings of the 10th International
Conference, Riga, Latvia, pp. 163-171. ISBN
978-9984-818-34-4, 2013.
[14] Sorana, D. B., Lorentz, J., Adriana, F. S., Radu, E. S. and Doru,
C. P., “Pearson-Fisher Chi-Square Statistic Revisited“, Journal
of Information,Vol. 2: pp.528-545; doi: 10.3390/info2030528;
ISSN 2078-2489, 2011.
[15] Downward, L. B., Lukens, W. W. & Bridges, F., “A Variation
of the F-test for Determining Statistical Relevance of Particular
Parameters in EXAFS Fits“, 13th International Conference on
X-ray Absorption Fine Structure, pp.129-131, 2006.
[16] LING, M., “Compendium of Distributions: Beta, Binomial,
Chi-Square, F, Gamma, Geometric, Poisson, Student's t, and
Uniform“, The Python Papers, Source Codes 1:4, 2009.
[17] Liang, J., “Testing the Mean for Business Data: Should One
Use the Z-Test, T-Test, F-Test, The Chi-Square Test, Or The
P-Value Method? “, Journal of College Teaching and Learning,
Vol. 3, pp. 7, doi:10.19030/tlc.v3i7.1704, 2011.