Student[Statistics][TwoSamplePairedTTest] Overview
overview of the Two Sample Paired T-Test
Description
Example
The two sample paired t-test is used to test if two samples come from the same population by comparing paired values between the samples. A paired samples t-test is based on a matched-pairs sample, where each value in one sample has a similar match in terms of other measured variables in the other sample.
The Student[Statistics][TwoSamplePairedTTest] function computes the paired t-test upon two datasets X1 and X2. This means that every entry of the population of X1 is related to an entry of the population of X2; in the samples, X1[i] is related to X2[i]; and the procedure tests whether the mean of the population of differences between related pairs, X1[i] minus X2[i], is equal to a value beta, under the assumption that these differences are normally distributed.
Requirements for using the Two Sample Paired T Test:
In order for the data in the two samples to be paired the sample size of the two samples must be the same.
The paired differences are assumed independent.
The paired differences are all assumed to follow identical normal distribution (same mean and variance).
The formula is:
$T=\frac{\left(\mathrm{Mean}\left(Z\right)-\mathrm{\beta}\right)\sqrt{N}}{s}$
where $Z$ is the sample whose data is generated as the difference between the data in each pair (${Z}_{i}={X}_{i}-{Y}_{i}$ ; $X$ is Sample One and $Y$ is Sample Two), $\mathrm{\beta}$ is the test value of difference between the means, $s$ is the sample standard deviation of $Z$, and $N$ is the sample size of $Z$.
$T$ follows Student's T distribution with $N-1$ degrees of freedom.
A group of 10 children aged 7 was randomly selected to watch an episode of a TV series called WashBill RoundShirt. After five years these 10 children were gathered to watch that episode again. The number of times when they laughed out loud during the episode was recorded:
child1
child2
child3
child4
child5
child6
child7
child8
child9
child10
age 7
8
5
6
10
7
11
12
age 12
9
13
Now the WashBill RoundShirt directors want to use the data to show that children from these two age groups consider WashBill RoundShirt equally funny.
Determine the null hypothesis:
Null Hypothesis: Children aged 7 and children aged 12 think WashBill RoundShirt is equally funny.
$\mathrm{\beta}=0$ in this case
Substitute the information into the formula:
$Z=X-Y$, where $X$ is the sample of group of children aged 7, $Y$ is the sample of group of children aged 12.
Thus $Z=\left[1\,0\,1\,\mathrm{-1}\,1\,0\,\mathrm{-1}\,\mathrm{-2}\,3\,0\right]$
$t=\frac{\left(0.2-0\right)}{\left(\frac{1.39841}{\sqrt{10}}\right)}=0.452267$
Compute the p-value by checking the table of Student's T distribution with 9 degrees of freedom:
$p-\mathrm{value}=\mathrm{Probability}\left(\left|T\right|>0.452267\right)=\mathrm{Probability}\left(T<-0.452267\right)+\mathrm{Probability}\left(T>0.452267\right)=0.66178$, $T\u02dc\mathrm{StudentT}\left(9\right)$
Draw the conclusion:
This statistical test does not provide enough evidence to conclude that the null hypothesis is false, so we fail to reject the null hypothesis.
See Also
Student[Statistics][TwoSamplePairedTTest]
Download Help Document