My dependent variable is vote choice between the three major Canadian federal parties, using Liberals as the reference point. I have three different lines of code drawn up to study the effect of visible minority status, region, and the interaction between visible minority status and region, on vote choice:
mlogit pes_vote_choice i.vis_min_status##i.region, baseoutcome(0) rrr
mlogit pes_vote_choice i.region i.vis_min_status#i.region, baseoutcome(0) rrr
mlogit pes_vote_choice i.region i.vis_min_status i.vis_min_status#i.region, baseoutcome(0) rrr
Now, this is still a preliminary analysis as I plan to add more control variables and interaction terms. But keeping that aside for now, just how do I interpret each of these results? What exactly is the difference between these three different regression results?
For example, looking at the output for the line 2 code, the interaction term Non-European descent#Atlantic has an RRR > 1 (2.14) and is statistically significant (P > |z| = 0.050). Does this mean that, relative to voters of European descent in the Atlantic, voters of non-European descent in the Atlantic are marginally more likely to vote NDP over Liberal, all else held equal? If so, clearly the same conclusion does not hold for the Conservatives in the Atlantic given the non-statistically significant result (P > |z| = 0.117), yet for the first line of code output, the result for the Conservative IS statistically significant (P > |z| = 0.033). Why does the first line of code output and second line of code output yield different results? Also, the third line of code includes the individual effect for visible minorities, and the output for that does appear completely identical to the first line of code output. If I am right with my interpretation of the Atlantic example from the second line of code output, can I apply the same reasoning to the output from the first and third lines of code? For the first and third line of code, does Non-European#Atlantic still illustrate the simple effect of a European descent/non-European difference on NDP/Liberal AT region = Atlantic? Because I would prefer a way to easily interpret that, which I am not sure the first and third lines do, but please let me know what their results mean exactly. Thank you.