Tuesday, August 04, 2009

A word about t-test (statistics)


A real scholar is a one with solid understanding of the topic and can address the concept in various terms. Just like a good teacher can address concepts in students' terms.
The t-test by virtue of its definition is a measurement of the separation of means. By adding the statistical flavor, we then asked the question of under what situation should we consider the means are well separated or, in other word, that the means are really not equal. Or in the statistical sense that the separation are statistically significant.

Like a lot of tests in statistics, the probability of the separation is too complicate to provide intuitive understanding of the problem. As in most of these kind of cases, a tabulated approach is used. In this case, a t-value is used to map to the probability. The t-probability table, however, usually only listed for positive t values which correspond to cases when the mean is greater than the referenced mean. Depend on the hypothesis of interest, researcher may need to included the probability for the corresponding negative t-value.

A good understanding of the underlying question that the t-test trying to address enables people to communicate with various terminology. A real scholar is a one with solid understanding of the topic and can address the concept in various terms. People with shaky ground are those that hanging to the jargon. Accusing people of alien-tongue is just a sign of it.

No comments: