This paper considers the method of data analysis for a pharmacological experiment with multiple dose groups, in which a response variable is observed repeatedly at several time periods. The inappropriateness of conventionally used statistical methods such as the t-test, Dunnett's multiple comparison procedure and analysis of variance method is pointed out. We recommended that the dose-response relationship should be estimated as a response function of time incorporating the dose effect as parameters. We also propose to adopt a strategy that the researcher tests some hypotheses on the parameters in the response function appropriate to the objective of the experiment after estimating parameters of the response function based on the observed data. A numerical example applying the strategy to real data is presented.
