Can I ask, where did you get the function to perform this test?

I usually use Tukey post hoc test in SPSS, you can run it indicates which one of your modalities had significant effect in contrast to other modalities

A friend gave me the function: 

lm1

It's  not 12, It's VI number 2

Thank you for links and informations about this test. 

Dear colleagues. I have analogues problem. Could i estimate effects by Scheirer-Ray-Hare test but  perfom pairwise comparison by Kruskal-Wallis or Mann-Whitney with following correction (for example Bonferroni)

Hi,

I'm not sure but I think that you can use this type of pairwise comparison using an iterative method like a Bonferroni-Holm correction. With this correction you can estimate the main effect with the Sheirer Ray Hare and if there is a significant effect you can do a sub--plan analysis (with Kruskall-Wallis for example) with a downgrading of the significance threshold. On an iterative way, if this second analysis is significative you can make a third analysis with another downgrading of the significance threshold.

I'm not sure but I hope that will help you.

Dear Romain! Thank you!

If i understand right. I do Sheirer Ray Hare analyzis, then i perform Kruskall-Wallis for parwize comparison, and then - Bonferroni-Holm correction ?

No I would say that you can apply a Bonferroni-Holm correction on all of your analyses. You have to begin with the analysis of your main hypothesis (in your example the Sheirer Ray Hare) and then you can perform other tests (as a Kruskall-Wallis for example) with a downgrading of the significant threshold for each analysis you add.

I don't know if I'm clear but you can look at this link for more information about the Bonferroni-Holm correction.

Have a nice day.

http://www.statisticshowto.com/holm-bonferroni-method/

Dear Romain!

Im sorry, but i want precise

I have two insecticides and two temperature (predictors),correspondantly - 4 rows of variables (mortality rates). I estimated effects of insecticides, temperature and their interactions by Sheirer Ray Hare. But! How could i estimate differenses between all 4 treatments (post-hoc). What corrert method for pair-wise comparison? May be Mann-Whitney followed by Bonferroni-Holm correction? 

The Bonferroni-Holm correction is apply to any test to reduce the significant threeshold of your analyses. It's not a test but a correction used in a test

The post hoc analysis the most indicate in that case seems to be a Tukey (as aforesaid by Salvatore).

I hope that will help you. I don't think that a Mann-Whitney is appropriate unfortunately...

Good luck and I'm sorry to can't be able to help you more. You will perhaps find more information by asking your question in the forum.

Ok thank you very much. I did not see the negation in your answer, sorry.

Have a nice day.

Dear colleagues. We thank all for comments.  We performed Sheirer Ray Hare test as reviewer required. Then we performed  Mann-Whitney pairwise comparison followed by Bonferroni correction for posttest. Referee was satisfied and paper was accepted (Journal of invertebrate pathology).

As a follow-up to this discussion. 

I did find an "official" description of this  test:  Sokal and Rohlf, 1995.  And I did add it to my R package  ( https://www.rdocumentation.org/packages/rcompanion/versions/1.9.1/topics/scheirerRayHare ).

However, in contemporary times, the test doesn't seem to be held in the highest regard.

I didn't do anything in depth, but from some comparisons I made to approprate parametric models on a few data sets, the Scheirer–Ray–Hare test is less likely to find the interaction effect significant.

Some examples are at: http://rcompanion.org/handbook/F_14.html .

https://www.youtube.com/watch?v=N729aMGIUOk

Why not try "Aligned Ranks Transformation ANOVA"?!

Following the comment by Zhenxing Zhang , I would also recommend using alternative approaches rather than the Scheirer-Ray-Hare test. Ordinal regression, aligned ranks transformation anova, or an appropriate generalized linear model. Good implementations of these approaches have appropriate post-hoc tests.

Dear Sal Mangiafico , i can't find post hoc method for interaction effect. Which seems strange, because we use the method itself mainly to determine interactions. Do you know any R code that allows one to do it? Regards

@TomaszRak , are you using the S-R-H test ? If so, I don't know of any published method for post-hoc testing. Using Dunn test on the interaction might make sense. That is,constructing a single factor for all the interaction levels. ... But the other approaches mentioned have post-hoc methods.

Tomasz Rak (previous reply).

Sal Mangiafico , yes, i'm using S-R-H test. I understand that there are other methods, but we are talking about using this particular test here. I found it odd that it didn't have a proper post-hoc method for interactions, since its main use is precisely the study of interactions. The examples you indicate and which can be found on the web look twofold: either there is no interaction at all, or with a significant interaction, the graph is simply analyzed (which, as we know, is not the correct approach).

Of course, one can perform post-hoc testing, for example, with the Dunn-Bonferroni method for all subgroups, but it can still be a criticized approach, it does not contain simple effects... and we still do not get an answer to the question: why does the test for non-parametric interaction have no test procedure post hoc at all.

BTW, did you notice that someone also recommended your post about you mistagging me? ;)