Time series

#1
Hi all,
In my experiment I collected data (dependent variable) from linear ccd camera every seconds during the period of 90 seconds. This procedure I have repeated five times. The design of the study requires comparison of the control and experimental group (each group size is set to 26), so I am in doubt whether to use ANOVA with repeated measures like 90x5 matrix, bootstrapping or some other test?
Many thanks for considering my question.
Andreja Vujanac
 
Last edited:
#2
Most likely, the assumption of sphericity will be violated. So repeated measures ANOVA is not the answer. You would need to use models for panel data. Such models would include autoregressive terms to capture correlation over time. They would also include fixed effects or random effects to capture unique properties of each of 26 cameras in each group. Group membership would be coded with a binary predictor: 1 for experimental, 0 for control.

You would need to play with many models and choose the best one using diagnostic tests and model selection criteria.