Hello,
In a dataset collected in 11 different cities, I was investigating the effect of environmental attitudes on car modal share (the % of car use in your everyday travel). The code I used for the analysis is
The resulting output shows that, yes, there is considerable variation in the effect of attitudes on modal share. Furthermore, the output revealed that the coefficients were strongly ordered by GDP. When I export the coefficients to excel using
and then making a simple scatterplot with GDP (PPP), the r2 is 0.88.
Although this provides a very clean graphical illustration, I'm convinced there is a better way to integrate GDP into the analysis. Some sort of interaction seems likely, but my knowledge of multilevel models is not sufficient to determine what solution is appropriate. Any suggestions for how to proceed are much appreciated.
Best,
Lars Even Egner
In a dataset collected in 11 different cities, I was investigating the effect of environmental attitudes on car modal share (the % of car use in your everyday travel). The code I used for the analysis is
Code:
mixed carfreq_share age female income pn_sum || city: pn_sum, ml variance cov(unstructured)
predict ebs ebi, reffects
replace ebs = _b[carfreq_share:pn_sum] + ebs
graph dot (mean) ebs, over (city, sort(ebs)) scheme (s1mono) plotregion(style(none)) title("Env. att. effect on car modal share indifferent cities") ytitle("Estimated coefficients [EB], %pts")
Code:
bysort city: sum ebs
Although this provides a very clean graphical illustration, I'm convinced there is a better way to integrate GDP into the analysis. Some sort of interaction seems likely, but my knowledge of multilevel models is not sufficient to determine what solution is appropriate. Any suggestions for how to proceed are much appreciated.
Best,
Lars Even Egner
