Comparing one group over 4 timepoints?

#1
What test and posttest would be most appropriate for a sample size of 10 assuming equal variances and normal distribution at a baseline and 3 timepoints? I need to compare each timepoint against the baseline. The parameter is body temperature.
 

hlsmith

Omega Contributor
#2
I don't think you can really assume normality given the sample size. I would like to see the results of a normality test and Q_Q plot to believe that, and that all four groupings meet that assumption?

I would use the wilcoxon with rank test, and you may want to use a smaller alpha value given multiple testing.
 
#3
Usually, you can't reject normality for small sizes even if the distribution is not normal ...not enough power to reject the normality ...
 

hlsmith

Omega Contributor
#4
@obh correct. My oversight in writing too quickly. This is not too far removed from getting 5 heads in a row on a fair coin.

But I would argue that it is difficult to accept such a small sample as meeting normality assumptions. The data generating process may result in an underlying normal random variable, but a small sample of it may not meet modeling or test assumptions.
 
#5
Yes hlsmith , you can't really determine if the data is normally distributed with a small sample. but T test is robust to the normality assumption if the data is reasonably symmetric. Is 10 too small even if data is reasonably symmetric. ? (you may argue what is reasonably symmetric ...)
 

hlsmith

Omega Contributor
#7
You could post the respective histograms for data to help guide the discussion.

Side note, if you used Wilcoxon test the correction for repeated comparisons would still be applicable and need to be addressed.
 
#9
Hi R.
Histogram - the number of observations per body temperature ranges.
can you please show us the histogram per each time point?