I'm running multiple regressions based on my IV of living in a district that varyies in percent latino. I use a range of 20 to 70% latino in increments of 5. Each % is a different dummy variable. So regression 1 tests living in a 20% latino district, regression 2 tests living in a 25% latino district, and so on, up to 70%.
When running my regression I use an interaction term of %latino and latino. For example, my first regression uses the following code:
*REGRESSION WITH 20%*
eststo clear
reg votedrecent i.latino##i.twty_pct_latino [w=vosuppwt], r
eststo
areg votedrecent i.latino##i.twty_pct_latino youth lowinc lowed [w=vosuppwt], absorb(year) r
eststo
areg votedrecent i.latino##i.twty_pct_latino youth lowinc lowed [w=vosuppwt], absorb(metarea) r
eststo
*RESULTS TABLE*
esttab, compress
esttab using "C:\Users\Renee.Sanjuan\Documents\reg20.rtf", replace onecell se
My result is the following table:
(1) | (2) | (3) | |
votedrecent | votedrecent | votedrecent | |
1.latino | -0.281*** (0.00373) |
-0.198*** (0.00362) |
-0.192*** (0.00373) |
1.twty_pct_latino | -0.0569*** (0.00224) |
-0.0701*** (0.00213) |
-0.0345*** (0.00794) |
1.latino#1.twty_pct_latino | 0.0708*** (0.00489) |
0.0766*** (0.00466) |
0.0672*** (0.00484) |
youth | -0.199*** (0.00183) |
-0.199*** (0.00186) |
|
lowinc | -0.105*** (0.00236) |
-0.107*** (0.00239) |
|
lowed | -0.196*** (0.00179) |
-0.194*** (0.00182) |
|
_cons | 0.539*** (0.00113) |
0.682*** (0.00133) |
0.669*** (0.00306) |
N | 373855 | 373855 | 373855 |
* p < 0.05, ** p < 0.01, *** p < 0.001
I was advised to use the margins command to determine the actual effect from my model. However, when attempting the following code I get an error that reads "at level for factor twty_pct_latino not present in estimation".
margins votedrecent, at(twty_pct_latino = (20(5)70))
What's the best way to interpret this data and ensure I'm describing the results accurately?
Thank you!
**** I am using Stata 15.1
0 Response to Measuring effects from interaction term
Post a Comment