What is the appropriate test for comparing means? I have a case and control group, both of which had a blood sugar level tested multiple times over 48 hours. I took the mean of the levels for each person and then calculated the mean of the means (with an SD etc) to compare the two groups. I am not sure if this is a valid method for comparison and whether another method is more appropriate? Thanks
 A: If you are strictly looking to compare means you can use an unpaired t-test.
This assumes that both independent populations are close to normally distributed and have somewhat equal variances.  The variance assumption loosens as n gets bigger.
You'll need to specify what you expect to see in the case group:


*

*higher blood sugar

*lower blood sugar

*any difference in blood sugar


You will use this to decide which one or two tailed test to use based on this.
Per Glen_b's advice perhaps you should create some plots of the data over time and see if there are some other interesting effects at play as well.
A: The answer to your question depends on the question you are trying to answer! The method you described may work well for some questions and not for others. For example, your method assumes
That blood sugar levels at different times have equal importance. For example, a high (or low) blood sugar at hour 2 is of the same importance as a high (or low) blood sugar at hour 37. 
If this is in line with the question you are trying to answer, great. If not you may consider (as Glen suggested) a model for your collection of variables, in particular one that allows you to estimate the effect of time (regression could help you do this).
Another aspect that will help you answer your question is how you compared the means. You mentioned that you used a SD, etc. This sounds like a good start, but you might want to do a more in depth check of your method. If you haven't already, I would consider looking up the assumptions and procedure for two-sample t-tests, which will provoke good questions like did you pool the standard deviations or not? Was there good evidence to do so, etc.
