Using summary measures for repeated observations
Matthews, JNS et al (1990) describe various summary measures to represent repeated measures. It is meant to be used when the repeated measures are measured over some continuum such as time points or levels of severity. The table summarising when and what to use is reproduced below.
Type of data |
Question to be answered |
Summary Measure |
||||
Peaked |
Is the overall outcome value the same in different groups? |
Overall mean (equal time intervals); Area Under Curve (unequal time intervals) |
||||
Peaked |
Is the maximum (minimum) response different between groups? |
Maximum (minimum) value |
||||
Peaked |
Is the time to maximum (minimum) response different between groups? |
Time to Maximum (minimum) response |
||||
Growth |
Is the (linear) rate of change of the outcome different between groups? |
Regression coefficient |
||||
Growth |
Is the eventual value of the outcome variable different between groups? |
Final value of outcome or difference between first and last values or percentage cnhange between first and last |
||||
Growth |
Is the response in one group delayed relative to the other? |
Time to reach a particular value (for example, a fixed percentage of baseline) |
||||
The use of summary measures for repeated observations is also supported by Lancaster (2009) who actually references Matthews et al. (1990) in this regard and Der and Everitt (2009, p.224). Lancaster also mentions Guyatt's responsiveness statistic.
References
Der G and Everitt BS (2009) A handbook of statistical analyses using SAS. Third Edition. Chapman and Hall:London.
Lancaster GA (2009) Statistical issues in the assessment of health outcomes in children: a methodological review. Journal of the Royal Statistical Association A 172(4) 707-728.
Matthews JNS, Altman DG, Campbell MJ and Royston, P (1990) Analysis of serial measurements in medical research. BMJ 300 230-235.