Dear all,
In the course of my thesis I’m investigating voting behavior in favor of right-wing populist parties in western Europe, using the last two waves of the European Social Survey (ESS7 and 8).
Beside economic variables I’ll include variables measuring attitudes towards immigration as well as some variables measuring a concept of “populism” (although not flawless with the given data).
I’ve used factor analysis for different items of immigration and what you could call “populism”, to investigate which of these variables load for the same factor as well as to create indices afterwards.
Two questions appeared to me while doing this.
The variables of the example range from 1 to 4 (zuw1-zuw3) and from 0 to 10 (immig1-immig3)
Index creation through factor analysis:
Index creation by myself:
Regression:
Thanks for help!
In the course of my thesis I’m investigating voting behavior in favor of right-wing populist parties in western Europe, using the last two waves of the European Social Survey (ESS7 and 8).
Beside economic variables I’ll include variables measuring attitudes towards immigration as well as some variables measuring a concept of “populism” (although not flawless with the given data).
I’ve used factor analysis for different items of immigration and what you could call “populism”, to investigate which of these variables load for the same factor as well as to create indices afterwards.
Two questions appeared to me while doing this.
- First, I’m confused how the results of my logistic regression differ including the generated indices instead of using an additive index created by myself out of the same variables. E.g. for immigration the odds-ratio differ from 1.592*** (additive index) to 2.412*** (generated index). Is this just “normal” or have I overlooked something?
- Second, to decide how many factors I’ll include I would like to use parallel analysis (in addition to the Kaiser criterion and looking at a screeplot). Searching the net guided my to the “fapara” program. The problem is, that I can’t find it anymore. Typing “search fapara” “findit fapara” doesn’t help. I’ve tried using the “paran” command, but it doesn’t really satisfy me. Any suggestions?
The variables of the example range from 1 to 4 (zuw1-zuw3) and from 0 to 10 (immig1-immig3)
Index creation through factor analysis:
Code:
factor zuw1 zuw2 zuw3 ///
immig1 immig2 immig3, blanks(0.4)
estat kmo
screeplot, yline(1)
rotate, promax normalize blanks (0.4)
estat common
predict indxzuw indximmig
rotate, clear
Index creation by myself:
Code:
egen indxzuw2 = rowtotal(zuw*), miss replace indxzuw2 = (indxzuw2/12)*100 egen indximmig2 = rowtotal(immig*), miss replace indximmig2 = (indximmig2/30)*100
Regression:
Code:
eststo clear
eststo: quietly logit wahlat alter geschl test_ess ///
abl3m einkq i.oekus i.bildung ///
indxzuw2 indximmig2 rlskala beweu ///
indxvpolnat indxdemozu indxvpolinat indxpolverdr ///
[pw=pspwght]
vif, uncentered
esttab, se mtitles wide scalars(r2_p) eform (1 1 1 1 1)
Thanks for help!
Comment