Hello all,
I will perform t-test for independent sample in EXCEL. There are two options t-test for equal variances and t-test for unequal variances.
i want to know the difference between two. I also want to know for which type of data these two are applicable.
Thanks
Dear Shanti Verma ,
The two versions of the test run different algorithms. There is a rule of thumb for the assumption of equality of variances that their ratio should be less than 4:1 (and thus the ration of standard deviations should be less than 2:1) for you to select the equal variances assumed version.
There is a formal test for selecting between these versions called Levene's test for equality of variances. However, it is not available in Excel. It is run automatically in SPSS when you run an independent t-test.
Dear Shanti Verma ,
The two versions of the test run different algorithms. There is a rule of thumb for the assumption of equality of variances that their ratio should be less than 4:1 (and thus the ration of standard deviations should be less than 2:1) for you to select the equal variances assumed version.
There is a formal test for selecting between these versions called Levene's test for equality of variances. However, it is not available in Excel. It is run automatically in SPSS when you run an independent t-test.
The t-test for unequal variances uses the Welch-Satterthwaite correction. When the variances are equal it gives essentially the same results as the equal variance test, so a common view is always to use the unequal variance option or always to use the unequal variance version if the results differ. This is better, in my view, than using something like the Levene test (which has robustness and power issues and addresses fundamentally the wrong question - whether the variances differ rather than whether the unequal variances have an impact on the outcome of the t test).
- Badges
- Science topic