The study endpoint is a cholesterolemia monitored for the course of a lifetime. In this experiment we have 2 groups: a group of individuals with familial hypercholesterolemia and a group of individuals without the disease. The example below is different in respect to the first one. Example #2: the main effects that can be interpreted In this case only the interaction can be interpreted, namely, the value of one of the main effects must be interpreted in light of the value of the other main effect. The type of diet, that is, determines a blood pressure but only after a certain age. The same is true for the effect of a type of diet: it may be significant but only after a certain age (in fact, until about forty years, two groups have equal blood pressure, thanks to the compensation mechanisms of the youth). Age affects the value of pressure in one of two groups and not in both of them. For this reason, the main effect “age” cannot be interpreted. In this case the main effect “age” may be significant because your test will consider the average effect of time on both groups, and the contribution in this case of the “high calorie diet” group makes the test significant.
The arterial pressure will rise only for one of two categories (high-calorie diet), while for people who follow a normal calorie diet style, it will remain stable for entire life (or will increase slightly). Does it make sense in this case to interpret the main effects? No. Let’s say you get both the main effects and the interaction significant.
Two main effects are age and type of diet (normal or high calorie). We analyze the value of an average blood pressure in respect to age in two groups: subjects that follow a high-calorie diet and subjects that follow normal diet style. Example #1: main effects that cannot be interpreted In the first example the rule appears to be correct.
In my opinion the opposite is true: when a significant interaction emerges from the analysis of the data, it is absolutely necessary to describe and give an interpretation of what the main effects tell us. When we analyze data, one of the main rules that we learned is that we should not interpret the main effects when their interaction is statistically significant.Īctually this rule is, in some cases, an inconvenient simplification.
0 kommentar(er)
