3
$\begingroup$

I have 3 different treatments with the observations of 6 different repetitions. However, all these different treatments (groups) have different numbers of patients. The first group has 26 patients, the second group has 28 patients and the third group has 19 patients. Thus, I can say that I have unequal sample sizes for two-way repeated measures anova.

When I do the analysis by using SPSS, it calculates the sum of squares and degrees of freedom by using the minimum sample size of the third group, which is 19.

Solving the problem and doing the comparisons by hand is time consuming. I would like to know how to handle this kind of problem, which is unequal sample sizes in each groups, by using the computer. Is there any other program, (SAS or R?) that deals with that, apart from SPSS, which obviously can not?

$\endgroup$
0

3 Answers 3

1
$\begingroup$

Look into linear mixed effects modelling, it can handle missing repeats which the standard anova linear modelling approach cannot.

$\endgroup$
3
  • $\begingroup$ I have no idea how to use it and its property about handling the two-way repeated measures and main effects and interaction term. Can you be more specific or can you give some papers for these models handling unequal sample sizes in repeated measures (not unequal or missing repetitions)? $\endgroup$
    – ARAT
    Commented Aug 13, 2013 at 15:20
  • $\begingroup$ Is this even an issue of missing data? It just sounds like Uneven sample sizes. $\endgroup$
    – Behacad
    Commented Jun 24, 2014 at 4:31
  • $\begingroup$ @Oceansoul linear mixed models have many advantages over ANOVA in its various forms. See, for example, Gelman's A psychology researcher asks: Is Anova dead?. For an intro to rANOVA —> mixed models see Repeated Measures ANOVA, R.I.P.?. $\endgroup$
    – Alexis
    Commented Feb 17, 2015 at 18:22
1
$\begingroup$

GraphPad Prism can handle one-way ANOVA in this situation. It cannot handle missing observations for any patient, but has no trouble with different number of patients in different treatment groups. Try the free demo.

$\endgroup$
1
$\begingroup$

SAS proc mixed can to this. See their documentation. Unequal and unknown covariance matrices as well as unequal sample sizes are supported there. (I would recommend UN as covariance structure unless you are sure about a different structure, see the repeated statement.)

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.