lineair regression/binary logit/ordered logit

[Steven Somers]

Dear Stata users,

I am a beginner in Stata and working on an econometric analysis in which I model

life satisfaction as a function of a set of explanatory variables in 6 selected countries.

Data is used from ESSround1.

I am trying to build a correct lineair, binary logit and ordered logit model.

Please have look at my do file.

*Selecting countries

keep if cntry=="ES" | cntry=="DE" | cntry=="NL" | cntry=="SI" | cntry=="PL" | cntry=="BE"

*Recoding

tab tvtot
replace tvtot =. if tvtot>7
recode tvtot (2 3 =2) (4 5 =3) (6 7 = 4), gen(tvtot_x)
tab tvtot tvtot_x
label define tvtot_x 0 "No TV at all" 1 "Less than 0,5 h TV" 2 "More than 1.5-2.5 h TV" 3 "More than 1.5-2.5 h TV" 4 "More than 2.5 h TV"
label values tvtot_x tvtot_x
fvset base 1 tvtot_x
tab tvtot_x

tab wkhtot
replace wkhtot=. if wkhtot > 120
generate wkhtot_sq = wkhtot^2
tab wkhtot_sq
label variable wkhtot "Working hours"
label variable wkhtot_sq "Working hours, squared"

lookfor "income"
tab hinctnt, nolab
replace hinctnt=. if hinctnt > 13 
generate income = ln(hinctnt)
label variable income "ln(household income, PPP)"
tab income

lookfor "age"
tab agea
replace agea=. if agea > 95
label variable agea "Age"
generate agea_sq = agea^2
label variable agea_sq "Age, squared"
tab agea_sq

lookfor "gender"
tab gndr, nolab
replace gndr=. if gndr==9

lookfor "citizen"
tab ctzcntr, nolab
replace ctzcntr=. if ctzcntr > 2
recode ctzcntr (2 = 1) (1 = 2), gen(citizen)
tab ctzcntr citizen
label define citizen 1 "Foreigner" 2 "Citizen of country"
label values citizen citizen
label variable citizen "Citizen of the country"
tab citizen

lookfor "partner"
tab partner marital
tab partner marital, nolab
replace partner=. if partner == 9
replace marital=. if marital > 5
tab marital partner
generate living=.
replace living=1 if partner==1 & marital ==1
replace living=2 if partner==2 & marital ==1
replace living=3 if partner==1 & marital ==2
replace living=4 if partner==2 & marital ==2
replace living=5 if partner==1 & marital ==3
replace living=6 if partner==2 & marital ==3
replace living=7 if partner==1 & marital ==4
replace living=8 if partner==2 & marital ==4
replace living=9 if partner==1 & marital ==5
replace living=10 if partner==2 & marital ==5
tab living
label define living 1 "Married, living with partner" 2 "Married, not living with partner" 3 "Separated, living with partner" 4 "Separated, not living with partner" 5 "Dicorced, living with partner" ///
6 "Dicorced, not living with partner" 7 "Widowed, living with partner" 8 "Widowed, not living with partner" 9 "Never been married, living with partner" 10 "Never been married, not living with partner"
label values living living
tab living

lookfor "children"
tab chldhm, nolab
replace chldhm=. if chldhm > 3
recode chldhm (2 = 1) (1 = 2), gen(child)
tab chldhm child
label define child 1 "No children at home" 2 "Children at home"
label values child child
label variable child "Are children living at home?"
tab child

lookfor "education"
tab edulvla
tab edulvla, nolab
replace edulvla=. if edulvla > 6
tab edulvla

lookfor "mainact"
tab mainact
tab mainact, nolab
replace mainact=. if mainact >9

lookfor "domicil"
tab domicil
tab domicil, nolab
replace domicil=. if domicil > 6


*lineair regression

keep if essround==1
regress stflife i.tvtot_x wkhtot wkhtot_sq income agea agea_sq i.gndr i.citizen i.child i.edulvla i.mainact i.living i.domicil


lookfor "life"
tab stflife
tab stflife, nolab
replace stflife=. if stflife > 11

global ylist stflife
global xlist i.tvtot_x wkhtot wkhtot_sq income agea agea_sq i.gndr i.citizen i.child i.edulvla i.mainact i.living i.domicil

*logit model

logit $ylist $xlist [pw=dweight*pweight]

logit $ylist $xlist [pw=dweight*pweight], or

*ordered logit model

ologit $ylist $xlist [pw=dweight*pweight]

ologit $ylist $xlist [pw=dweight*pweight], or

Now I want to make a new logit model for ESSround5 so

I can compare it with ESSround1. Problem is that 2 variables have different names in ESSround5.

Here's the recoding for 'education' for round5.

replace edulvla=1 if (edulvlb==0 | edulvlb==113) & essround==5 
replace edulvla=2 if (edulvlb==129 | edulvlb==212 | edulvlb==213 | edulvlb==221 | edulvlb==222 | edulvlb==223) & essround==5
replace edulvla=3 if (edulvlb==229 | edulvlb==311 | edulvlb==312 | edulvlb==313 | edulvlb==321 | edulvlb==322 | edulvlb==323) & essround==5
replace edulvla=4 if (edulvlb==412 | edulvlb==413 | edulvlb==421 | edulvlb==422 | edulvlb==423) & essround==5
replace edulvla=5 if (edulvlb==510 | edulvlb==520 | edulvlb==610 | edulvlb==620 | edulvlb==710 | edulvlb==720 | edulvlb==800) & essround==5
replace edulvla=55 if edulvlb==5555 & essround==5
replace edulvla=77 if edulvlb==7777 & essround==5
replace edulvla=88 if edulvlb==8888 & essround==5
replace edulvla=99 if edulvlb==9999 & essround==5

There's a second problem for the marital status.

ESSround1: variables partner and marital (see do file)

ESSround5: variables icpart1 and marsts


How can I make my variables to be comparable with each other? How can I set up my new logit model for round5?

Please feel free to adjust my do file and/or give me some explanations.

Any corrections or suggestions are much appreciated!

Kind regards,
Steven Somers

[Maarten Buis]

'These are questions that can only be answered by studying the documentation of the data, in particular the exact question asked and the exact answer categories and how these answer categories were processed to produce your variable. How to ask about education in a cross national comparative manner is a complicated issue, I know of two dissertations that were written on how this was done in the ESS...