Filter löschen
Filter löschen

Different results from ranova and JASP

10 Ansichten (letzte 30 Tage)
Joshua Diamond
Joshua Diamond am 18 Mär. 2024
Beantwortet: Shivansh am 1 Apr. 2024
Dear all, I am struggling to make sense of ranova with multcompare vs JASP (a statistical program). They are providing different results.
What I have is a table where rows correspond to subjects, and columns correspond to conditions. I want to compare each condition to each other condition. Therefore I am using ranova and multcompare.
load('data.mat','resultsAll')
rTable = array2table(resultsAll);
varNames = strsplit(string(num2str(1:size(resultsAll,2))));
withinDesign = table(varNames','VariableNames',{'State'});
withinDesign.State = categorical(withinDesign.State);
conditionString = sprintf('%s-%s ~ 1',rTable.Properties.VariableNames{1},rTable.Properties.VariableNames{end});
rm = fitrm(rTable,conditionString,'WithinDesign',withinDesign);
fullTable = ranova(rm,'WithinModel','State');
disp(fullTable)
SumSq DF MeanSq F pValue pValueGG pValueHF pValueLB ________ __ _________ ______ __________ _________ __________ _________ (Intercept) 0.43612 1 0.43612 26.656 0.0066842 0.0066842 0.0066842 0.0066842 Error 0.065443 4 0.016361 (Intercept):State 0.43911 5 0.087821 20.282 3.2733e-07 0.0019177 0.00014406 0.010793 Error(State) 0.086601 20 0.0043301
c = multcompare(rm,'State');
disp(c);
State_1 State_2 Difference StdErr pValue Lower Upper _______ _______ __________ _________ ________ __________ _________ 1 2 0.3161 0.056406 0.02833 0.048612 0.58359 1 3 0.25331 0.059127 0.069544 -0.027085 0.5337 1 4 0.33228 0.056094 0.023317 0.066273 0.59829 1 5 0.3353 0.067885 0.043651 0.01338 0.65723 1 6 0.33858 0.068912 0.044434 0.011784 0.66537 2 1 -0.3161 0.056406 0.02833 -0.58359 -0.048612 2 3 -0.062793 0.040148 0.65386 -0.25318 0.12759 2 4 0.016182 0.0072343 0.37993 -0.018125 0.050488 2 5 0.019203 0.01907 0.89461 -0.07123 0.10964 2 6 0.022477 0.019813 0.84709 -0.071481 0.11643 3 1 -0.25331 0.059127 0.069544 -0.5337 0.027085 3 2 0.062793 0.040148 0.65386 -0.12759 0.25318 3 4 0.078975 0.037339 0.42181 -0.098094 0.25604 3 5 0.081996 0.035042 0.3474 -0.08418 0.24817 3 6 0.085269 0.034084 0.30158 -0.076361 0.2469 4 1 -0.33228 0.056094 0.023317 -0.59829 -0.066273 4 2 -0.016182 0.0072343 0.37993 -0.050488 0.018125 4 3 -0.078975 0.037339 0.42181 -0.25604 0.098094 4 5 0.0030213 0.016544 0.99994 -0.075434 0.081477 4 6 0.0062946 0.017727 0.99861 -0.077772 0.090361 5 1 -0.3353 0.067885 0.043651 -0.65723 -0.01338 5 2 -0.019203 0.01907 0.89461 -0.10964 0.07123 5 3 -0.081996 0.035042 0.3474 -0.24817 0.08418 5 4 -0.0030213 0.016544 0.99994 -0.081477 0.075434 5 6 0.0032733 0.0025707 0.78816 -0.0089175 0.015464 6 1 -0.33858 0.068912 0.044434 -0.66537 -0.011784 6 2 -0.022477 0.019813 0.84709 -0.11643 0.071481 6 3 -0.085269 0.034084 0.30158 -0.2469 0.076361 6 4 -0.0062946 0.017727 0.99861 -0.090361 0.077772 6 5 -0.0032733 0.0025707 0.78816 -0.015464 0.0089175
On the other hand, JASP gives very different results for a repeated measures ANOVA.
I get the same F value of 20.282.
But the p-values in multiple comparisons is very different.
I am using all the default parameters of JASP.

Antworten (1)

Shivansh
Shivansh am 1 Apr. 2024
Hi Joshua!
It looks like you are comparing the results from MATLAB and JASP. As you have mentioned, the "f-value" and "df" are similar in both cases.
For the pairwise comparison using "multcompare", I can observe that the mean difference is also consistent across the cases. The difference in standard error will lead to a difference in p-values and as a result, lower and upper bounds will also be different. There might be the case that the implementations in MATLAB and JASP use slightly different approaches for pooling variance or handling within-subject correlations, especially in the context of a repeated measure, affecting "StdErr", which in turn affects p-values and confidence intervals.
You can refer to the following link to know more about the "ranova" and "multcompare" functions in MATLAB:
I hope it helps!

Kategorien

Mehr zu Text Analytics Toolbox finden Sie in Help Center und File Exchange

Tags

Produkte


Version

R2023b

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by