As you know, the Student t-test can be used only to compare two groups. When you have k groups you can't use a t-test for each couple because, if you have fixed alpha=0.05 for each t-test, there is a probability of k*0.05 to find a difference when it isn't (Bonferroni inequality).
The Holm-Sidak test is a step-down "recursive reject", because it applies an accept/reject criterion on a sorted set of null hypothesis, starting from the lower p-value and going up to the acceptance of null hypothesis.
For each comparison, the alpha value is set according to Sidak correction of Bonferroni inequality.
Statistics Toolbox is required because this function use TCDF function.
Yes it is possible but quite incovenient, I'll have to look through this script to find the correct variables to output.
My point is simply that this script doesn't seem quite finished, the variables are there they are just not being output at the moment...
I thank you for the script, but I can't understand why it doesn't output the info everyone will clearly want.
I don't agree. If you are so able to write a script, it would not to be so difficult to write 1 line of code to obtain an output struct. You can modify my script as you want: it is open source.
This code is useless if you want to incorporate it as part of a script... it doesn't really 'output' anything - just some info in the command screen.
This file is similar to multcompare matlab function, but provides much more clear outputs.
I think that my explanation is very simple and if "someone is crap at stats" he never will search this file.
sorry but this does not explain anything to someone who is crap at stats.
Changes in description
logical indexing improved
Changes in help section
Correction to avoid error generation when more comparisons have the same p-value.
Logical indexing implemented
Error handling and calculation improvements
Download apps, toolboxes, and other File Exchange content using Add-On Explorer in MATLAB.