Is there a reason why you cannot formulate a single regression model: Y = b0 + b1X1 + b2X2 + b3X3 + b4X1*X3 + b5X2*X3 ??
It, ostensibly, would seem to me, to be much easier to work with - statistically speaking.
Dear
I'm wondering how I should interpret my hypotheses.
The first one is "the relationship between X1 and Y increases as X3 increases" and the second one is exactly the same but with X2, so "the relationship between X2 and Y increases as X3 increases"
I have used a regression analysis with the models:
Y = β0 + β1X1 + β2X3 + β3X1*X3
Y = β0 + β1X2 + β2X3 + β3X2*X3
However by doing the analysis in SPSS the significance for the first hypothesis is p=0.682 and p=0.934 for the second (This was after centering the variables to counter collinearity and coding X3 into a continious variable. It was a categorical variable before.).
All of this was done after asking a professor from the university. Still I'm wondering if this means the relationship between X1/X2 and Y declines when X3 increases? How should I interpret this? Am I using the right method?
I also have a third hypothesis about which X (X1 or X2) has the greatest influence on Y. How could I analyse this? Could I use a Shapley Owen Decomposition or eta-squared? Or is there an easier method?
Kind regards
Is there a reason why you cannot formulate a single regression model: Y = b0 + b1X1 + b2X2 + b3X3 + b4X1*X3 + b5X2*X3 ??
It, ostensibly, would seem to me, to be much easier to work with - statistically speaking.
Tweet |