Sample size calculation is a key component in designing experiments. It has economical and ethical implications in experimental design. Insufficient sample size can lead to a lack of power in detecting departures from the null hypothesis, whereas a large sample size may expose subjects to unnecessary risks. Despite its importance, only few modifications have been made to the way statisticians calculate sample size even if the complexity of the research questions has increased. In this talk, I will present three research projects motivated by my consulting practice. The first two projects deal with developing new methodology to calculate the sample size of, first, an experiment that requires a mixture of discrete and continuous correlated random variables and, secondly, when the data to be analyzed has functional form. The third project presents a counter-intuitive result regarding the "optimal" number of groups that yield the total sample size necessary to achieve certain power at a given significance level in a one-way ANOVA design.