I performed the one way ANOVA and multiple comparisons to deal with my data. But what's the statistical method of the function "multcompare"? Tukey's or Sidak's or ...

I performed the one way ANOVA and multiple comparisons to deal with my data. But what's the statistical method of the function "multcompare"? Tukey's or Sidak's or ...

Answer 1

[William Rose]

@杰,

[Edit: Correct spelling mistakes.]

multcompare() uses the Tukey-Kramer (also known as honestly significant difference) test by default. You can specify other statistical tests with the optional "CriticalValueType" parameter.

For example,

c = multcompare(stats,"CriticalValueType","bonferroni")

See here for a full list of options. Good luck with your analysis.

[William Rose]

Example:

[Edit: I posted my initial comment before it was complete.]

Example:

• Create simulated data set with three groups of 20 normally distributed values. The mean of group 2 is higher than groups 1 and 3.
• Analyze the data with anova1().
• If anova p<0.05, perform multple comparisons. Use the default statistics (Tukey-Kramer). Then repeat using Bonferroni statistics.
• The confidence intervals (columns 3 and 5 of ans) and the p-values (column 6 of ans) for the multiple comparisons are slightly different for the different statistical methods.

rng(42);                                % for reproducibility of this example
y=randn(20,3)+repmat([0 1 0],20,1);     % simulated data, three groups of 20
[p,~,stats]=anova1(y,[],'off');         % perform anova
if p<0.05       
    multcompare(stats,Display='off')
    multcompare(stats,CriticalValueType='bonferroni',Display='off')
end

ans = 3×6

1.0000 2.0000 -1.6377 -0.8384 -0.0391 0.0378 1.0000 3.0000 -0.7865 0.0128 0.8121 0.9992 2.0000 3.0000 0.0519 0.8512 1.6505 0.0344

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

ans = 3×6

1.0000 2.0000 -1.6577 -0.8384 -0.0190 0.0432 1.0000 3.0000 -0.8065 0.0128 0.8322 1.0000 2.0000 3.0000 0.0319 0.8512 1.6705 0.0392

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>