So I am running partial correlations in SPSS, and each time I add a new variable that I am controlling for, the number of participants in the analysis steadily decreases. I analyzed frequencies and have no missing values, however. In fact, I know that I don't have missing values because every value entered is an average. Does anyone know why SPSS is excluding people from the correlation when they are not missing any values? Interestingly, when I look at frequencies after intentionally deleting a subjects values, the frequency analyses still returns an output that indicates still no values are missing. This is so bizarre. Has anyone encountered this before? Advice on how to rectify the situation?
Since partial corr reports df, not number of cases, and the control variables consume degrees of freedom, adding a control variable will reduce that number.