Dear all
I'm working with three level nested data:
Individuals (childnum6), nested within...
Nuclear families (linked by nfamid), nested within...
Extended families (linked by efamid)
*A sample of my data is attached below*
My simplified baseline OLS regression to estimate a LPM is:
To control for extended family fixed effects, I de-mean the variables (denoted with d_) with respect to extended family means and run:
Question 1: Since I de-mean to manually remove extended family fixed effects, I need to adjust the standard errors. As I cluster at the nuclear family level, I can not use -, dof()- to do this and instead need to calculate: true variance = report variance*(report DoF/true DoF) and se = sqrt(variance). At the moment I'm doing this with a calculator. Is there is a way to calculate this within Stata using matrices? I'm not proficient enough with Stata commands to achieve this myself. Any help would be massively appreciated.
I augment the fixed effects model with an IV component. The variable d_spfamsize is used as an instrument for endogenous chtotal6. Since both these variables are at the nuclear family level, I need to run the first stage of 2SLS at the nuclear family level (using iv_d_spfamsize) and the second stage at the individual level - i.e., need to do it manually.
Question 2: Now I need to adjust standard errors for the de-meaning and the use of predicted values in the IV procedure. Again, any idea how to do this?
Any help would be hugely appreciated.
Best
Owen
I'm working with three level nested data:
Individuals (childnum6), nested within...
Nuclear families (linked by nfamid), nested within...
Extended families (linked by efamid)
*A sample of my data is attached below*
My simplified baseline OLS regression to estimate a LPM is:
Code:
reg chmarried6 chtotal6 childnum6 chage6 feduc, cluster(nfamid)
Code:
reg d_chmarried6 d_chtotal6 d_childnum6 d_chage6 d_feduc, cluster(nfamid)
I augment the fixed effects model with an IV component. The variable d_spfamsize is used as an instrument for endogenous chtotal6. Since both these variables are at the nuclear family level, I need to run the first stage of 2SLS at the nuclear family level (using iv_d_spfamsize) and the second stage at the individual level - i.e., need to do it manually.
Code:
*First Stage reg d_chtotal6 iv_d_spfamsize d_childnum6 d_chage6 d_feduc predict d_chtotal6_hatz, xb bysort nfamid: egen d_chtotal6_hat = mean(d_chtotal6_hatz) *Second Stage reg d_chmarried6 d_chtotal6_hat d_childnum6 d_chage6 d_feduc, cluster(nfamid)
Any help would be hugely appreciated.
Best
Owen
Code:
* Example generated by -dataex-. For more info, type help dataex
clear
input long efamid float nfamid byte childnum6 float(chmarried6 chtotal6 chage6 feduc spfamsize d_chmarried6 d_childnum6 d_chtotal6 d_chage6 d_feduc d_spfamsize iv_d_spfamsize)
1200002 4 1 0 3 45 13 8 -.5 -1 0 3.333332 0 0 0
1200002 4 2 . 3 42 13 8 . 0 0 .3333321 0 0 .
1200002 4 3 1 3 38 13 8 .5 1 0 -3.666668 0 0 .
1200004 7 1 1 2 42 20 3 0 -.5 0 2.5 0 0 0
1200004 7 2 1 2 37 20 3 0 .5 0 -2.5 0 0 .
1200005 8 1 1 2 49 14 2 .5 -.5 0 .5 0 0 0
1200005 8 2 0 2 48 14 2 -.5 .5 0 -.5 0 0 .
1200007 12 1 1 3 38 16 . 0 -1 0 -2.799999 -.5 . .
1200007 12 2 . 3 35 16 . . 0 0 -5.799999 -.5 . .
1200007 12 3 1 3 . 16 . 0 1 0 . -.5 . .
1200007 13 1 1 3 46 17 4 0 -1 0 5.200001 .5 0 0
1200007 13 2 1 3 43 17 4 0 0 0 2.2000008 .5 0 .
1200007 13 3 1 3 42 17 4 0 1 0 1.2000008 .5 0 .
1200009 15 1 1 9 50 13 9 0 -3.25 3 3.583332 -.5 0 0
1200009 15 2 1 9 48 13 9 0 -2.25 3 1.583332 -.5 0 .
1200009 15 3 1 9 47 13 9 0 -1.25 3 .58333206 -.5 0 .
1200009 15 4 1 9 46 13 9 0 -.25 3 -.4166679 -.5 0 .
1200009 15 5 1 9 45 13 9 0 .75 3 -1.416668 -.5 0 .
1200009 15 6 1 9 44 13 9 0 1.75 3 -2.416668 -.5 0 .
1200009 15 7 1 9 44 13 9 0 2.75 3 -2.416668 -.5 0 .
1200009 15 8 1 9 41 13 9 0 3.75 3 -5.416668 -.5 0 .
1200009 15 9 1 9 33 13 9 0 4.75 3 -13.416668 -.5 0 .
1200009 16 1 1 3 60 14 . 0 -3.25 -3 13.583332 .5 . .
1200009 16 2 1 3 53 14 . 0 -2.25 -3 6.583332 .5 . .
1200009 16 3 1 3 46 14 . 0 -1.25 -3 -.4166679 .5 . .
1200010 18 1 0 2 34 17 . -.5 -.5 0 2 0 . .
1200010 18 2 1 2 30 17 . .5 .5 0 -2 0 . .
1200011 19 1 1 2 45 12 3 0 -.5 0 1 0 0 0
1200011 19 2 1 2 43 12 3 0 .5 0 -1 0 0 .
1200012 21 1 1 4 47 12 3 .25 -1.5 0 3.75 0 0 0
1200012 21 2 1 4 45 12 3 .25 -.5 0 1.75 0 0 .
1200012 21 3 0 4 43 12 3 -.75 .5 0 -.25 0 0 .
1200012 21 4 1 4 38 12 3 .25 1.5 0 -5.25 0 0 .
1200013 23 1 0 3 44 17 2 -.6666667 -1 0 2 0 0 0
1200013 23 2 1 3 42 17 2 .3333333 0 0 0 0 0 .
1200013 23 3 1 3 40 17 2 .3333333 1 0 -2 0 0 .
1200015 28 1 0 2 42 12 4 0 -.5 0 1.5 0 0 0
1200015 28 2 0 2 39 12 4 0 .5 0 -1.5 0 0 .
1200016 29 1 1 3 52 13 4 .6666666 -1 0 2 0 0 0
1200016 29 2 0 3 50 13 4 -.3333333 0 0 0 0 0 .
1200016 29 3 0 3 48 13 4 -.3333333 1 0 -2 0 0 .
1200017 31 1 0 2 30 14 5 -.6 -.8 -.5 -8.200001 1 1.5 1.5
1200017 31 2 0 2 30 14 5 -.6 .20000005 -.5 -8.200001 1 1.5 .
1200017 32 1 1 3 48 12 2 .4 -.8 .5 9.799999 -1 -1.5 -1.5
1200017 32 2 1 3 47 12 2 .4 .20000005 .5 8.799999 -1 -1.5 .
1200017 32 3 1 3 36 12 2 .4 1.2 .5 -2.2000008 -1 -1.5 .
1200018 33 1 0 3 49 12 . -.3333333 -1 0 0 0 . .
1200018 33 2 0 3 . 12 . -.3333333 0 0 . 0 . .
1200018 33 3 1 3 . 12 . .6666666 1 0 . 0 . .
1200019 35 1 1 4 44 18 3 .3333333 -1.1666667 1 6.833332 1 1.5 1.5
end
