crossover study, also referred to as a crossover trial, is a longitudinal study in which subjects receive a sequence of different treatments (or exposures). While crossover studies can be observational studies, many important crossover studies are controlled experiments, which are discussed in this article. Crossover designs are common for experiments in many scientific disciplines, for example psychology, education, pharmaceutical science, and medicine.

Randomized, controlled crossover experiments are especially important in health care. In a randomized clinical trial, the subjects are randomly assigned to different arms of the study which receive different treatments. When the randomized clinical trial is a repeated measures design, the same measures are collected multiple times for each subject. A crossover clinical trial is a repeated measures design in which each patient is randomly assigned to a sequence of treatments, including at least two treatments (of which one "treatment" may be a standard treatment or a placebo).

Nearly all crossover designs have "balance", which means that all subjects should receive the same number of treatments and that all subjects participate for the same number of periods. In most crossover trials, in fact, each subject receives all treatments.

Statisticians suggest that designs have four periods, a design which allows studies to be truncated to three periods while still enjoying greater efficiency than the two-period design.[1][2] However, the two-period design is often taught in non-statistical textbooks, partly because of its simplicity.

Thanks Fariborz for the explanation. My study design just had one group and all of the participants received the same treatment " education sessions" . How I deal with missing data multiple imputations or last observation carried forward?

Ghada, although a single arm (presumably pre- post-) design doesn't get a lot of attention, say in medicine, it is still sometimes used in pilot phase. Although LOCF still is widely used in medicine, statistically it has been pretty well discredited. As you could see, one obvious problem would be the variance alone. The more observations are missing as time goes on, the lower the variance becomes with respect to time because you keep replacing values with ones that have no variance with respect to time. Secondly, assuming that you have a basic understanding of multiple imputation, the LOCF observations do not include any sort of component of unreliability. Whereas an imputed value under MI includes both the plausible value component, in each data set there is an added error term to increase the standard errors to accommodate the fact that the imputed values are not as reliable as the observed ones. In LOCF this is not the case and the values carried forward are treated just the same as the observed ones. There's actually quite a bit written about this, which I cannot begin to capture here, but MI is certainly better than LOCF. Bob