I concur with the author’s assertion that in repeated measures design, dependencies are formed among data points; however, I would question the notion that this is always a “complexity.” Although it may confuse analysis, it also enriches the study design and discourages individual differences. When the same group of participants was to be measured over again, then we lessen the variability arising from differences between people, which can actually reduce the statistical power of the test. This is by no means an exhaustive list of situations in which it helps to have a firm grasp on the concept, but consider just one example. In behavior analysis or health research, we often measure how well someone is doing, how stressed they are, or how much better they are getting at more than one point in time. This way,y you will be able to utilize the test to verify that observed changes are attributed only to intervention and not random variation or participant differences. Hereby, the material does an interesting job in showing us the importance of acknowledging dependencies in longitudinal data and using appropriate statistical techniques, having that in mind as opposed to assuming that repeated observations are indeed independent.

I would like you to explain to me the criteria used by researchers when deciding which test is better paired t-test or the independent t-test, in situations when samples are connected, but it is not too clearly explained, for example, when participants have certain characteristics in common but are technically individual personalities. Is this still something that can be referred to as “paired,” or should it be handled as independent? From the professional viewpoint, it is very important to know which type of t-test should be used so that you can interpret data correctly. For example, in health and behavioral research, when deciding on the most appropriate test for us, it may be a question of whether or not a certain treatment is considered effective. Misuse of the test can hurt by providing wrong conclusions that affect real-world interventions.

Reading this section, one point I got from it most was on how you should not use a tool before you understand the assumptions that come with it. The text also explains to non-statisticians why parametric tests, such as the t-test, are not always the best choice, and this conjured up some reflective thoughts on how much real-world testing there is for normality in the data. It also raised some inquiries on the Kolmogorov-Smirnov test. I want more explanation of its difference with other tests for normality, like the Shapiro-Wilk test, and when it may be used in preference to the other.