ranova(rm) ranovatbl contains only NaN values and zero's

ranova(rm) outputs a ranova table containing only NaN values and zero's:

ranovatbl =

  3×8 table

                                            SumSq    DF    MeanSq     F     pValue    pValueGG    pValueHF    pValueLB
                                            _____    __    ______    ___    ______    ________    ________    ________

    (Intercept):ReactionTime                  0      0      NaN      NaN     NaN        NaN         NaN         NaN   
    ReactionTimes_Indicator:ReactionTime      0      0      NaN      NaN     NaN        NaN         NaN         NaN   
    Error(ReactionTime)                       0      0      NaN                                                       

This is my rm fit:

allRTSLong.ReactionTimes_Indicator = categorical(allRTSLong.ReactionTimes_Indicator);
Meas = table([1]','VariableNames',{'ReactionTime'});
rmRTS = fitrm(allRTSLong, 'ReactionTimes ~ ReactionTimes_Indicator', 'WithinDesign', Meas)
[ranovatbl] = ranova(rmRTS)

This is what the first 10 rows of allRTSLong look like:

>> allRTSLong(1:10, :)

ans =

  10×2 table

    ReactionTimes_Indicator    ReactionTimes
    _______________________    _____________

          Visual                   0.3745   
          Auditory               0.057861   
          Audiovisual             0.44841   
          Visual                  0.33798   
          Auditory                0.20852   
          Audiovisual             0.16943   
          Visual                   0.3462   
          Auditory                 0.1226   
          Audiovisual             0.66347   
          Visual                  0.36488   

I've tried making ReactionTimes_indicator catagorical, tried various different inputs instead of 'Meas'. Nothing has worked so far. I've looked everywhere on the internet but i couldn't find any solutions.

Also, all data are from one mouse. Am i even using the right test?

Answer 1

[Jeff Miller]

> 'WithinDesign', Meas

This doesn't look right with Meas only having one possibility. Try removing that.

> all data are from one mouse. Am i even using the right test?

I don't think so. The analysis you are doing is for situations where you have many randomly selected subject/mice and multiple measures on each one. If you have only one mouse, then the trials are the only source of random variation in the data set so the trials are the "subjects" and it is essentially a between-Ss design (i.e., each trial is only tested in one condition). But the results of the ANOVA will only tell you about systematic effects for this one mouse, and you will have no statistical basis for generalizing to other mice. For that, you would need multiple mice.

[Thomas61197]

Thank you!