Oneway Bonferroni

#1
Hello,

I am trying to compare variables mean and std deviation across three samples. Is oneway bonferroni test is the best way to do it? If yes, when reporting P value in descriptive statistics table, which p value should be included? ( there are p values for each row column mean comparison and another one for F test) Thanks
 

obh

Active Member
#2
Hi Econ,

I didn't hear about "oneway Bonferroni test", but Bonferroni correction for multiple comparisons -in your case if you running multiple t-tests to compare each pair of averages you must reduce the significant level of each test to get the desire to combine significant level.

I can only guess ...you that you run a combined test of one way ANOVA and multiple t-tests with Bonferroni correction.
If so the F tests the following H0: all the averages are equal. and t tets all the i-j combinations like H0 mean(i)=mean(j)

Can you please paste the results? what software do you use?

I think the most common is using the one way ANOVA (if mett the assumptions ...), and only if you reject the H0 you can run Tukey-HSD for each pair to find the reason for the ANOVA rejection. Tukey-HSD for equal group size or Tukey-Kramer (similar test) for unequal group size)

You can run the Levene's test to compare the standard deviations (probably with median Brown-Forsythe)
 

hlsmith

Not a robit
#3
Also you dont have to report pvalues. Report estimates with corrected confidence intervals! Everyone thinks you need pvalues, but you don't - estimates are way more informative.
 

obh

Active Member
#4
Also you dont have to report pvalues. Report estimates with corrected confidence intervals! Everyone thinks you need pvalues, but you don't - estimates are way more informative.
You can report both p-values (and the effect size) and the confidence intervals.
 

obh

Active Member
#5
It seems that Statistics is a "tradition science", for example, why do the z-table or any table still exist? We should send the statistics' tables to the same place we send the fax...