I am estimating a discrete choice model with the help of cox regression in SPSS.
I would like to interact two continuous predicting variables.
The main effect of one of the ineracting variables is not displayed in the output( degreee of freedom reduced because of constant or linearly dependet covariates). Can I still interpret the interaction term? (The interaction is significant)
I assume the variable is omitted because it is constant per stratum.
The overall question is how consumers choose between different options within a choice set of 3 items.
One sub-question is if the price has an effect of the choice probability of an item. I also predicted that this effect is moderated by the range of the price of the given choice set. (i.e. the difference between the smallest and the highest price within the choice set). Therefore within one choice situation for 1 participant (1stratum) the variable “price_range” is constant. Since participants became different choice sets this variable varies over the strata.
The setup of my research was that participants were confronted with different kind of choice sets. Each choice consisted of 3 bottles. Participants were asked to pick one out of the 3 bottles shown to them. In total I had 12 different choice sets that varied in size of the bottles and price. Participants were randomly confronted with 2 of these choice sets. Now I want to estimate the influence of the attributes of a the choice options (size and price) on its choice probability. For that reason I chose cox regression, because this is (as far as I understood) the only way to do this kind of conditional logit in SPSS.
Below you find an extract how the data is set up. I set it up in a long format to be able to run the cox regression. The example shows 1 respondent making choices in 2 independent choice sets. So the stratum variable basically only groups each choice set represented to a partipant. As you can see, the range is constant for one choice set. I am not interested in the effect of the range on choice probability, only on the interaction effect of price_range and price.