# T Test

There are two kinds of statistics; descriptive and inferential. Descriptive statistics merely describes the sample data that has been collected. Inferential statistics gives us a better understanding of the results and allows us to make inferences beyond the sample data. The T Test falls in the second category. The T Test is a comparison of the variance between the groups and the variance within the groups. The test tells us if the differences that have been shown by the means are reliable, and whether they are likely to occur the next time when the same experiment is carried out. It also tells us whether there is a real reason for the differences in the samples or it was just a matter of chance.

There are three different kinds of T Tests. The first type of test is conducted by taking independent samples. The mean averages of two different groups are taken and compared. Besides being known as the independent samples test, it is also known as between samples and unpaired samples test. The comparison of two different drugs or a comparison of two different beverages would fall under this category. The second type of test is known as the paired samples test or repeated measures test. This is when the mean of the same sample is tested at two different times. The testing of a group of subjects before and after taking a drug would fall under this category. This test also allows us to compare the scores of individual subjects at different times. The third type of test is called one sample test. In this version, a hypothetical mean is assumed, and the observed mean of the group is compared to it.

While this test is very useful, there are certain limits to the results derived from it. The generalisations made on the basis of the result can only be done on samples similar to the one in the experiment. If the testing of a drug was done on a group of males over the age of fifty, then the results of the test would not apply to males below that age group. The distribution of scores should not be skewed; all the numbers should be more or less close to the mean. The sample populations that are being compared should roughly have the same number of subjects. Comparing a large population with a small population leads to unreliable results.