I collected data from 50 patients at 3 different time points and performed analysis using 3 different methods. What would be a more appropriate test to compare the measurements between time and methods, a two-way repeated measures ANOVA or a mixed ANOVA, and why? I am using the SPSS definition so apologies if these tests have other generic names. Thanks!
6
-
$\begingroup$ Any suggestions? $\endgroup$– Xe MCommented Jan 29, 2020 at 16:02
-
1$\begingroup$ Please provide more information. What is your research question? How did the results differ between the two approaches? I don't use SPSS but I believe that the mixed-effects ANOVA assumes you have at least one within-subjects factor and one between-subjects factor. You've only described a single within-subjects factor, so it is not clear how you ran that model. $\endgroup$– dbwilsonCommented Jan 30, 2020 at 14:03
-
$\begingroup$ I am measuring a physiological marker, e.g. heart rate, and want to determine if the measurements differ between time points and methods. The latter should be the two factors, no? My knowledge says that since the same patients were used for all time points and methods, I should use the 2-way repeated measures (within-within) ANOVA. But it has been suggested to me to use the mixed ANOVA, but can't see why. $\endgroup$– Xe MCommented Jan 31, 2020 at 17:04
-
$\begingroup$ Yes, this is a two-way repeated-measures ANOVA (within-within). The term "mixed ANOVA" often gets used in the context of repeated measures data to handle the clustering of observations within people. I don't know enough about SPSS's mixed to know if you can specify two within factors and not between factors. If you can, you should be the same results as with a repeated-measures ANOVA. Do you have multiple observations within each time point and method within person? $\endgroup$– dbwilsonCommented Feb 1, 2020 at 1:36
-
$\begingroup$ Ok I got it, thanks. There is one observation for each time point and method within person, so in my case 9 observations in total per person. $\endgroup$– Xe MCommented Feb 4, 2020 at 10:20
|
Show 1 more comment
1 Answer
$\begingroup$
$\endgroup$
This can then be run as a two-way repeated-measures ANOVA in the GLM procedure if you have the data set up with one case per person and the outcome values in nine different variables. This is sometimes called multivariate data structure, or wide data.
It can also be handled in the MIXED procedure if the data are set up with nine cases per person, sharing the same value on an ID variable, and using two index variables for the two factors, with a single dependent variable. This is sometimes referred to as the univariate data setup, or as narrow data.
The GLM procedure can also be used with the narrow setup and explicit naming of the subject ID as a random factor, but I'd suggest using MIXED if you're going to go at it from that angle.
