bootstrap oneway - ANOVA multiple comparisons

[Wendy Jane]

Hi,

I am trying to use bootstrapping for my ANOVA, however I cannot figure out how to do multiple comparsions with bootstrapping as it comes up with an error:

sysuse cancer 
bootstrap, reps(2000) seed(2): oneway _t drug, bonferroni

Is there a way to overcome this?

Thank you very much,
Wendy

[Joseph Coveney]

You need to tell bootstrap what it is that you're bootstrapping, that is, it needs an expression list:

bootstrap exp_list [, options eform_option] : command

[Anne Ahrens]

Hello,

I want to bootstrap my mean since data is non-normal distributed, to run a Sidak multiple-comparison test.
I have tried the following code, but get
'r(mean)' evaluated to missing in full sample
r(322);

Code example:
use https://www.stata-press.com/data/r18/apple
set seed 1234
bootstrap weight_2=r(mean), reps(1000): oneway treatment weight , sidak



Thank you for your help

/Anne

[Carlo Lazzaro]

Anne:
as -oneway- does not return -r(mean)-, Stata is obviously right in throwing -r(322)-:

. use "C:\Program Files\Stata17\ado\base\a\auto.dta"
(1978 automobile data)

. oneway price rep78, sidak

                        Analysis of variance
    Source              SS         df      MS            F     Prob > F
------------------------------------------------------------------------
Between groups      8360542.63      4   2090135.66      0.24     0.9174
 Within groups       568436416     64      8881819
------------------------------------------------------------------------
    Total            576796959     68   8482308.22

Bartlett's equal-variances test: chi2(4) =  11.4252    Prob>chi2 = 0.022

                  Comparison of Price by Repair record 1978
                                   (Sidak)
Row Mean-|
Col Mean |          1          2          3          4
---------+--------------------------------------------
       2 |    1,403.1
         |      1.000
         |
       3 |    1,864.7    461.608
         |      0.993      1.000
         |
       4 |      1,507    103.875   -357.733
         |      0.999      1.000      1.000
         |
       5 |    1,348.5    -54.625   -516.233     -158.5
         |      1.000      1.000      1.000      1.000

. return list

scalars:
                  r(N) =  69
                r(mss) =  8360542.627898574
               r(df_m) =  4
                r(rss) =  568436416.2416667
               r(df_r) =  64
                  r(F) =  .235327431924253
           r(chi2bart) =  11.42516013277495
            r(df_bart) =  4

.

That said, the normality of the mean should not be an issue.
I addition, I would go -regress- that can handle heteroskedastcity and other relevant stuff that are not considered by -oneway-.

[Anne Ahrens]

Hi Carlo

Thank you for explaining why r(mean) did not work with oneway.

Do you have another suggestion for how I can obtain multiple-comparison test for non-parametrical data?

/Anne

[Felix Bittmann]

Well, I have absolutely no idea whether this is statistically sound, but you could bootstrap adjusted p-values (if this has any benefit for your research...).

clear all
use https://www.stata-press.com/data/r18/apple
expand 2         //Make sample bigger for bootstrap approach to run properly


cap program drop comparer
program define comparer, rclass
    reg weight i.treatment
    margins treatment, pwcompare(pveffects) mcompare(sidak)
    return scalar comp2_1 = r(table_vs)[4,1]
    return scalar comp3_1 = r(table_vs)[4,2]
    return scalar comp4_1 = r(table_vs)[4,3]
    return scalar comp3_2 = r(table_vs)[4,4]
    return scalar comp4_2 = r(table_vs)[4,5]
    return scalar comp4_3 = r(table_vs)[4,6]
end

comparer
return list



bootstrap r(comp2_1) r(comp3_1) r(comp4_1) ///
    r(comp3_2) r(comp4_2) r(comp4_3), reps(500): comparer
estat bootstrap, bc