Daniel - Repeated measures refers to measurements made repeatedly over time on the same individual. An internet search will help you get an R script to fit to your situation. IITA biometrician S Korie at Ibadan may help you with this.

Hi Daniel,

Subhash gave the correct answer in that repeated measures is on the same unit over time (whether that unit is an individual, group, town, country, etc.). However, I am not sure I understood your approach clearly. Are you saying that each chick is repeatedly measured, or are they measured once and then not remeasured? This may be the difference between repeated measures and repeated cross-sectional measurements (different chicks at each time point)?

Please clarify...

Ariel

Hallo Daniel,

As the previous responses indicate, sampling over time of distinct individuals without any repeated sampling from the same individual does not fall within the repeated measures model. As I understand your description, you therefore do not require a repeated measures analysis. Instead, an analysis of variance with the five treatments as categorical variable and time as covariate would be appropriate.

Depending on the form of changes over time, it will probably be necessary to include non-linear time terms (e.g. quadratic) in the model or (if a limited number of discrete sampling time points is involved) to include time as a categorical covariate.

I am not familiar with r packages: look for those with either ANCOVA or general linear models (GLM).

The repeated measures can be conducted in the R software with GLMM or GAMM model.

Hi there.

Thank all of you so far. First to clarify Ariel - I think Jeremy has already done it in his contribution, but for more clarity, each chick was sampled only once. The notion of repeatability was in each treatment group. Because of this, I also see Jeremy differ in opinion on using a repeated measures model. Concerning expected changes in the variable over time, much changes are expected with serum antibody titres (log transformed) over time following  vaccination- changes best expressed and compared in a quadratic model. Thanks a lot Jeremy and all of you.

It's worth noting that the issue of repeated measures has mostly to do with the independence of the measurements, a critical assumption in many statistical models. When repeated measures occur over time occur in the same individual, the within-subject correlations may violate the assumption of independence. This is true for repeated measures, but other sources of clustering can also lead to correlated errors (e.g. clustering by family, household, clinic, or hospital). This is the feature that dictates the use of statistical models to account for the clustering. For example, imagine if the vaccine response depended on the litter or housing status (shared cages). The assumption of independence may again be violated even with only one measurement per chick. Once this is answered, the question of which method would be appropriate can be addressed, as there are several options e.g. 1) whether to model the marginal effects by empirically adjusting the covariance matrix (GEE); 2) to explicitly model the covariance structure to account for the correlations (GLS); or 3) to rely on random effects to induce an appropriate correlation structure (mixed effect models). In the absence of a 'test for independence', subject matter expertise is crucial. 

Thanks a lot Atul. Subject matter expertise is in deed crucial and I will combine it with your suggestions.