Bonferroni Correction to ttest

[Lucas Kiely]

Is there a way of incorporating a Bonferroni Correction in the ttest command e.g.

  
ttest BRIER_WTN = BRIER_Rank

or do I need to manually apply the correction using the 0.05/number of comparisons to the p-values?

I have 3 comparisons

  
ttest BRIER_WTN = BRIER_Rank
ttest BRIER_WTN = BRIER_UTR
ttest BRIER_UTR = BRIER_Rank

So its not a huge ordeal

[Joseph Coveney]

. . . . do I need to manually apply the correction using the 0.05/number of comparisons to the p-values? I have 3 comparisons . . .So its not a huge ordeal

That would probably be your fastest, but if you wanted an official pretty printout, then you could do something like the following

rename (BRIER_WTN BRIER_Rank BRIER_UTR) (bri#), addnumber(0)
generate `c(obs_t)' row = _n
quietly reshape long bri, i(row) j(typ)

label define Types 0 BRIER_WTN 1 BRIER_Rank 2 BRIER_UTR
label values typ Types

xtreg bri i.typ, i(row) fe

pwcompare typ, pveffects  mcompare(bonferroni)

which would give identical results to your pairwise series of Student's t-tests with Bonferroni adjustment.

[Joseph Coveney]

. . . which would give identical results to your pairwise series of [paired] Student's t-tests with Bonferroni adjustment.

I take that back: xtreg , fe will pool the errors and so won't be exactly the same.

If you want to follow the three paired Student's t-tests exactly, then you could do the following.

generate byte k = 1

manova BRIER_WTN BRIER_Rank BRIER_UTR = k, noconstant

margins , pwcompare(pveffects) mcompare(bonferroni)

[Lucas Kiely]

Thank you, that worked well