Statistical analysis of longitudinal data where some data are missing
I have found a particular type of statistical model (*) particularly powerful for analysis of longitudinal data from animal experiments where there are occasional missing values (e.g., where an animal drops out of the study due to ill-health).
I have written a click-by-click guide to analysing longitudinal data using such models in SPSS. This manuscript has been submitted for publication to Nature Protocols.
There are some accompanying SPSS data files you can download to learn how to use these models to analyse longitudinal data from animals where some data are missing. To download these, Right Click and "Save As" the following link: You will then need to Unzip this to a new folder: we recommend using the free evaluation version of WinZip. Then use one of the following files: For SPSS v18, use "long_format.sav" and "short_format.sav". For earlier SPSS versions, import data using "Portable" files: "long_format.por" and "short_format.por" or import data from Excel files "long format.xls" and "short_format.xls" (for the latter, see also Read_me.doc).
The zipped file also contains SPSS Syntax files for analysing these data files using Repeated measures analysis of covariance (RM ANCOVA) ("RM_ANCOVA.sps") or using mixed models with Maximum Likelihood estimation and covariance structures of the following types: UN, CS, AR1. Also available is a Syntax file for analysing this data file with the CS covariance structure and REML estimation.
Click here to view additional statistical models resources.
(*) models with "general covariance structure of the residuals" where parameters are estimated using Restricted Maximum Likelihood Estimation. Don't be put off by the name; it's as straightforward as repeated measures ANOVA to do!