I have a panel dataset across three waves where the Hausman test has advised to use a fixed effects estimator.
I wish to look at the effect of local area unemployment (psum_unemployed_total_cont_y) on health behaviors (smoking = no_cigs_cons_more0_y).
I would like to include time-invariant effects (baseline numbers of offspring and level of education = own_education_y calt3_other_children_y0) so I opt for a Hybrid Model as described by Richard Williams here: https://www3.nd.edu/~rwilliam/Taiwan2018/Hybrid.pdf
This is as below:
The problem that I am facing is that I don't know which coefficient to report?
My earlier random effects regression was as below, which leads me to believe that I should report "mpsum_unemployed_total_cont_y" but I don't want to just report the random effects coefficient again (if this is what this is), I would like to report the hybrid coefficient, i.e. the result of considering both random and fixed effects!
I wish to look at the effect of local area unemployment (psum_unemployed_total_cont_y) on health behaviors (smoking = no_cigs_cons_more0_y).
I would like to include time-invariant effects (baseline numbers of offspring and level of education = own_education_y calt3_other_children_y0) so I opt for a Hybrid Model as described by Richard Williams here: https://www3.nd.edu/~rwilliam/Taiwan2018/Hybrid.pdf
This is as below:
Code:
generate insampm = 0
recode insampm 0 = 1 if has_y0_questionnaire==1 & has_y5_questionnaire==1 | has_y0_questionnaire==1 & has_y10_questionnaire==1 | has_y0_questionnaire==1 & has_y5_questionnaire==1 & has_y10_questionnaire==1 | has_y0_questionnaire==1 & cbmi_y5 !=. & has_y5_questionnaire==0 | has_y0_questionnaire==1 & cbmi_y10 !=. & has_y10_questionnaire==0 | has_y0_questionnaire==1 & cbmi_y5 !=. & has_y5_questionnaire==0 & cbmi_y10 !=. & has_y10_questionnaire==0 | has_y0_questionnaire==1 & cbmi_y5 !=. & has_y5_questionnaire==1 | has_y0_questionnaire==1 & cbmi_y10 !=. & has_y10_questionnaire==1 | has_y0_questionnaire==1 & cbmi_y5 !=. & has_y5_questionnaire==1 & cbmi_y10 !=. & has_y10_questionnaire==1
(insampm: 1362 changes made)
foreach var of varlist psum_unemployed_total_cont_y year maritalstatus_y medical_card_y employment_y age_y {
2. egen m`var' = mean(`var') if insampm, by (id)
3. }
(858 missing values generated)
(842 missing values generated)
(842 missing values generated)
(844 missing values generated)
(848 missing values generated)
(902 missing values generated)
foreach var of varlist psum_unemployed_total_cont_y year maritalstatus_y medical_card_y employment_y age_y {
2. gen d`var' = `var' - m`var' if insampm
3. }
(858 missing values generated)
(842 missing values generated)
(975 missing values generated)
(986 missing values generated)
(981 missing values generated)
(902 missing values generated)
xtreg no_cigs_cons_more0_y dpsum_unemployed_total_cont_y-dage_y mpsum_unemployed_total_cont_y-mage_y i.own_education_y i.calt3_other_children_y0 if insampm, cluster (current_county_y1) re robust
note: dage_y omitted because of collinearity
note: myear omitted because of collinearity
Random-effects GLS regression Number of obs = 1,114
Group variable: id Number of groups = 631
R-sq: Obs per group:
within = 0.0262 min = 1
between = 0.1809 avg = 1.8
overall = 0.1464 max = 2
Wald chi2(20) = 58403.84
corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0000
(Std. Err. adjusted for 29 clusters in current_county_y1)
-----------------------------------------------------------------------------------------------------------------------------
| Robust
no_cigs_cons_more0_y | Coef. Std. Err. z P>|z| [95% Conf. Interval]
------------------------------------------------------------+----------------------------------------------------------------
dpsum_unemployed_total_cont_y | -.0191279 .0168943 -1.13 0.258 -.0522402 .0139844
dyear | .0048042 .0034354 1.40 0.162 -.001929 .0115375
dmaritalstatus_y | -.0074616 .0192272 -0.39 0.698 -.0451461 .030223
dmedical_card_y | .1089937 .0437265 2.49 0.013 .0232914 .194696
demployment_y | -.001246 .0033976 -0.37 0.714 -.0079051 .0054131
dage_y | 0 (omitted)
mpsum_unemployed_total_cont_y | -.023618 .0048528 -4.87 0.000 -.0331293 -.0141067
myear | 0 (omitted)
mmaritalstatus_y | .078648 .0161179 4.88 0.000 .0470576 .1102385
mmedical_card_y | -.0056338 .0411143 -0.14 0.891 -.0862165 .0749488
memployment_y | -.0149283 .0078878 -1.89 0.058 -.030388 .0005315
mage_y | .0006872 .0028682 0.24 0.811 -.0049343 .0063088
|
own_education_y |
Some secondary school | -.2838905 .2284327 -1.24 0.214 -.7316103 .1638293
Complete secondary education | -.3103412 .2229378 -1.39 0.164 -.7472914 .1266089
Some third level education at college, university, RTC | -.3254963 .2247214 -1.45 0.147 -.7659421 .1149495
Complete third level education at college, university, RTC | -.4002531 .209009 -1.92 0.055 -.8099033 .0093971
|
calt3_other_children_y0 |
1 | .0395838 .0393032 1.01 0.314 -.0374491 .1166168
2 | -.0237422 .0418116 -0.57 0.570 -.1056915 .0582071
3 | -.0609639 .0360013 -1.69 0.090 -.1315251 .0095973
4 | .0535541 .1040949 0.51 0.607 -.1504681 .2575764
5 | .0848587 .1476662 0.57 0.566 -.2045616 .3742791
8 | -.1378285 .0502685 -2.74 0.006 -.2363529 -.0393041
|
_cons | .6280112 .209512 3.00 0.003 .2173753 1.038647
------------------------------------------------------------+----------------------------------------------------------------
sigma_u | .27673727
sigma_e | .2451217
rho | .56036107 (fraction of variance due to u_i)
-----------------------------------------------------------------------------------------------------------------------------
. estimates store randomlpm
The problem that I am facing is that I don't know which coefficient to report?
My earlier random effects regression was as below, which leads me to believe that I should report "mpsum_unemployed_total_cont_y" but I don't want to just report the random effects coefficient again (if this is what this is), I would like to report the hybrid coefficient, i.e. the result of considering both random and fixed effects!
Code:
. xtreg no_cigs_cons_more0_y psum_unemployed_total_cont_y i.calt3_other_children_y0 i.year i.own_education_y i.maritalstatus_y i.me
> dical_card_y i.employment_y i.ord_age_y if has_y0_questionnaire==1 & has_y5_questionnaire==1 | has_y0_questionnaire==1 & has_y10_
> questionnaire==1 | has_y0_questionnaire==1 & has_y5_questionnaire==1 & has_y10_questionnaire==1 | has_y0_questionnaire==1 & cbmi_
> y5 !=. & has_y5_questionnaire==0 | has_y0_questionnaire==1 & cbmi_y10 !=. & has_y10_questionnaire==0 | has_y0_questionnaire==1 &
> cbmi_y5 !=. & has_y5_questionnaire==0 & cbmi_y10 !=. & has_y10_questionnaire==0 | has_y0_questionnaire==1 & cbmi_y5 !=. & has_y5_
> questionnaire==1 | has_y0_questionnaire==1 & cbmi_y10 !=. & has_y10_questionnaire==1 | has_y0_questionnaire==1 & cbmi_y5 !=. & ha
> s_y5_questionnaire==1 & cbmi_y10 !=. & has_y10_questionnaire==1, cluster (current_county_y1) re robust
Random-effects GLS regression Number of obs = 1,114
Group variable: id Number of groups = 631
R-sq: Obs per group:
within = 0.0261 min = 1
between = 0.1812 avg = 1.8
overall = 0.1489 max = 2
Wald chi2(28) = .
corr(u_i, X) = 0 (assumed) Prob > chi2 = .
(Std. Err. adjusted for 29 clusters in current_county_y1)
-----------------------------------------------------------------------------------------------------------------------------
| Robust
no_cigs_cons_more0_y | Coef. Std. Err. z P>|z| [95% Conf. Interval]
------------------------------------------------------------+----------------------------------------------------------------
psum_unemployed_total_cont_y | -.0180869 .004859 -3.72 0.000 -.0276103 -.0085636
|
calt3_other_children_y0 |
1 | .0430077 .044525 0.97 0.334 -.0442597 .1302751
2 | -.0129626 .0387062 -0.33 0.738 -.0888254 .0629002
3 | -.0437057 .034716 -1.26 0.208 -.1117477 .0243364
4 | .068641 .0994023 0.69 0.490 -.1261839 .2634659
5 | .1617445 .1652178 0.98 0.328 -.1620766 .4855655
8 | -.1260235 .037658 -3.35 0.001 -.1998318 -.0522153
10 | 0 (empty)
|
5.year | .0561064 .01575 3.56 0.000 .0252368 .0869759
|
own_education_y |
No schooling | 0 (empty)
Primary school education | 0 (omitted)
Some secondary school | -.2338917 .2338158 -1.00 0.317 -.6921623 .224379
Complete secondary education | -.2680814 .2294854 -1.17 0.243 -.7178644 .1817017
Some third level education at college, university, RTC | -.2875848 .2313358 -1.24 0.214 -.7409946 .165825
Complete third level education at college, university, RTC | -.3700443 .2147431 -1.72 0.085 -.790933 .0508444
|
maritalstatus_y |
Cohabiting | .1500805 .0286762 5.23 0.000 .0938761 .2062849
Separated | -.0365335 .0722616 -0.51 0.613 -.1781637 .1050967
Divorced | .1687855 .1566259 1.08 0.281 -.1381957 .4757667
Widowed | .3805275 .2318141 1.64 0.101 -.0738198 .8348748
Single/Never married | .2121704 .0515474 4.12 0.000 .1111393 .3132015
|
medical_card_y |
Yes | .0614009 .0272528 2.25 0.024 .0079865 .1148154
|
employment_y |
Unemployed | .1442326 .123105 1.17 0.241 -.0970488 .3855141
Unable to work owing to permanent sickness or disability | .0265288 .0491606 0.54 0.589 -.0698242 .1228818
At school/student | -.0424344 .0463123 -0.92 0.360 -.1332048 .0483361
Seeking work for the first time | .1543783 .1988002 0.78 0.437 -.2352629 .5440195
Employed | -.021495 .0225956 -0.95 0.341 -.0657816 .0227916
Self Employed | -.0226135 .0337065 -0.67 0.502 -.0886771 .0434501
Wholly retired from paid work | -.07375 .0298647 -2.47 0.014 -.1322837 -.0152163
|
ord_age_y |
20-23 | -.0872846 .0915932 -0.95 0.341 -.2668039 .0922347
24-27 | -.0450939 .0737355 -0.61 0.541 -.1896128 .0994251
28-32 | -.0075531 .074343 -0.10 0.919 -.1532626 .1381564
33 + | -.0453948 .0805736 -0.56 0.573 -.2033162 .1125266
|
_cons | .5925997 .2514472 2.36 0.018 .0997723 1.085427
------------------------------------------------------------+----------------------------------------------------------------
sigma_u | .27722292
sigma_e | .24538799
rho | .56068997 (fraction of variance due to u_i)
-----------------------------------------------------------------------------------------------------------------------------
