t-Test Two-Sample

#1
Hi,
I perform t-Test Two-Sample in Excel using the Analysis Tool pack.
I used both methods: 1) Assuming Unequal Variances, 2) Assuming Equal Variances.
I get in both cases that they are significant.
How can it be? The H0 assumption is opposite in both cases. Isn't it?
Thanks,
Tamiri
 
#2
Hi Tamiri,

You need to run an F-test to determine if the variances are equal or not. That will determine which test to use. If your F-test is significant your variances are not equal and you need to use the t-test assuming unequal variances. The null hypothesis of both test is the same, I believe Excel calculates the p-value differently based on the distribution of the data in unequal vs. equal variances tests. Do some reading to find out how the tests are run so you are comfortable with your analysis.

Andrea
 

Dason

Ambassador to the humans
#3
How can it be? The H0 assumption is opposite in both cases. Isn't it?
No - the t-test has the same null and alternative hypothesis in both cases. The difference is what you assume about the data. When you assume equal variances you are saying that you are going to make the assumption that both groups have the same variance (not this is an assumption about the variance - not about the means). When you assume unequal variance you are actually just saying that the variance of the data for each group might be different.

Andrea said:
You need to run an F-test to determine if the variances are equal or not. That will determine which test to use. If your F-test is significant your variances are not equal and you need to use the t-test assuming unequal variances. The null hypothesis of both test is the same, I believe Excel calculates the p-value differently based on the distribution of the data in unequal vs. equal variances tests. Do some reading to find out how the tests are run so you are comfortable with your analysis.
I don't think they need to run the F-test as this is just homework and they are being asked to run both versions anyways. Also there are serious doubts as to whether doing a formal test of equal variance before deciding which test to use is actually a good idea.