Hi,
I noticed that the F value in F ratio test for equal variance in excel and F value in Levene's Test for Equality of Variance in SPSS is not similar for the same data.
Please find attached the results. What could be the reason for this difference in F value in these two different methods. Is it simply the difference in the formula to compute the F value?
Thanks and regards
Amit
If I understand the question, the ratio of variances, and Levene's test, while both testing HOV are different tests and should not be expected to yield the same effect. Levene's test is in favor as it is more robust to violation of normality.
I am not really trying to muddy the water, but to complete the picture, there is also another test of HOV due to Bartlett. The three give differing results, although in my experience, not usually substantively in respect to the null of HOV.
My best on your research,
John
Hi Daniel Wright ,
Please find attached the screen shot of the entire output obtained from the excel and spss for the F ratio and Levene's test for equality of variances.
I am also attaching the practice data set (the excel file and the SPSS file) from which this output is obtained.
Thanks and regards
Amit Basra
Thanks for attaching the data (note that the values here are different from in your original question … as far as I know all tests for the two group case will agree on which side of 1 the F value is on … i.e., which variance is larger). As John Morris said, there are lots of choices. I used your data to show some of these. SPSS default for Levene's uses mean centering, and this is not very robust. Any way, in the attach is doing a few of these tests in R.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS