Hi,I want to compare difference among eight groups using One-Way ANOVA in SPSS,but in the test of homogeneity of variances p<0.05,it is equal variances not assumed,so can I use One way ANOVA for further analysis? How about post hoc multiple comparisons?Thanks.
0
1 reply to this topic
#2
-
- Active Members
-
- 114 posts
Veteran
Posted 17 August 2009 - 04:19 PM
Will transforming the data, eg log scale, work? I’ve read that deviations from homogeneity can be tolerated if there are equal numbers of samples in each of your groups; but, like all statistics, there is no consensus. SPSS has a range of post hoc tests available when equal variances cannot be assumed (Tamhane's T2, Dunnett's T3, Games-Howell, and Dunnett's C) so presumably they recognise that we often have to precede the best we can with dodgy data.
Of those four post hoc tests above I would guess that Tamhane’s would be best for you, Dunnett’s tend to be used when you are comparing groups against a reference rather than against each other (but I’m pushing my knowledge of stats to the limit so double check
). In practice I tend to select a number of post hoc tests and make sure they all give similar answers or that I can give an explanation of any differences.
Of those four post hoc tests above I would guess that Tamhane’s would be best for you, Dunnett’s tend to be used when you are comparing groups against a reference rather than against each other (but I’m pushing my knowledge of stats to the limit so double check
). In practice I tend to select a number of post hoc tests and make sure they all give similar answers or that I can give an explanation of any differences.
