Note: This chapter uses a suite of commands, called spost, written by J. Scott Long and Jeremy Freese. The commands must be downloaded prior to their use, and this can be done by typing search spost in the Stata command line (see How can I use the search command to search for programs and get additional help? for more information about using search).
Table 7.1, page 191.
use https://stats.idre.ucla.edu/stat/stata/examples/long/tobjob2, clear
quietly reg jobcen fem phd ment fel art cit
listcoef, std
regress (N=408): Unstandardized and Standardized Estimates
Observed SD: .97360294
SD of Error: .8717482
-------------------------------------------------------------------------------
jobcen | b t P>|t| bStdX bStdY bStdXY SDofX
-------------+-----------------------------------------------------------------
fem | -0.13919 -1.543 0.124 -0.0680 -0.1430 -0.0698 0.4883
phd | 0.27268 5.529 0.000 0.2601 0.2801 0.2671 0.9538
ment | 0.00119 1.692 0.091 0.0778 0.0012 0.0799 65.5299
fel | 0.23414 2.469 0.014 0.1139 0.2405 0.1170 0.4866
art | 0.02280 0.789 0.430 0.0514 0.0234 0.0528 2.2561
cit | 0.00448 2.275 0.023 0.1481 0.0046 0.1521 33.0599
-------------------------------------------------------------------------------
quietly reg jobcen fem phd ment fel art cit if jobcen ~= 1
listcoef, std
regress (N=309): Unstandardized and Standardized Estimates
Observed SD: .77904266
SD of Error: .70309389
-------------------------------------------------------------------------------
jobcen | b t P>|t| bStdX bStdY bStdXY SDofX
-------------+-----------------------------------------------------------------
fem | 0.10145 1.187 0.236 0.0481 0.1302 0.0618 0.4744
phd | 0.29738 6.361 0.000 0.2758 0.3817 0.3540 0.9274
ment | 0.00078 1.273 0.204 0.0541 0.0010 0.0695 69.5468
fel | 0.14053 1.565 0.119 0.0662 0.1804 0.0850 0.4710
art | 0.00590 0.238 0.812 0.0142 0.0076 0.0182 2.4000
cit | 0.00210 1.271 0.205 0.0760 0.0027 0.0976 36.1466
-------------------------------------------------------------------------------
quietly tobit jobcen fem phd ment fel art cit, ll(1)
listcoef, std
tobit (N=408): Unstandardized and Standardized Estimates
Observed SD: .97360294
Latent SD: 1.21966
SD of Error: 1.087237
-------------------------------------------------------------------------------
jobcen | b t P>|t| bStdX bStdY bStdXY SDofX
-------------+-----------------------------------------------------------------
fem | -0.23685 -2.032 0.043 -0.1156 -0.1942 -0.0948 0.4883
phd | 0.32258 5.047 0.000 0.3077 0.2645 0.2523 0.9538
ment | 0.00134 1.514 0.131 0.0880 0.0011 0.0722 65.5299
fel | 0.32527 2.656 0.008 0.1583 0.2667 0.1298 0.4866
art | 0.03391 0.929 0.353 0.0765 0.0278 0.0627 2.2561
cit | 0.00509 2.057 0.040 0.1683 0.0042 0.1380 33.0599
-------------------------------------------------------------------------------
Figure 7.6, page 200.
Note: This graph was created in three steps since the prgen command after tobit produces expected values rather than probabilities.
Step 1: Create a data set for each of the categories, female fellow (ff), female nonfellow(fnf), male fellow(mf) and male nonfellow(mnf) where the level of PhD varies from 1 to 5 by .5 and the other variables in the model, ment, art and cit are held at there mean. The four data sets where created from the data set figure76 where for each respective data set, fem and fel where varied to define each case type.
Step 2: Run the full tobit model and use that model for out-of-sample prediction for the four data sets, ff, fnf, mf and mnf, where we predict the probability of being censored with the command predict varname, pr(.,1) as well as sorting the observations.
Step 3: Merge all the data sets together by the phd variable and graph the probability of censor to phd level.
Step 1.
clear input phd 1 1.5 2 2.5 3 3.5 4 4.5 5 end gen ment = 45.47058 gen art = 2.276961 gen cit = 21.71569 save figure76 file figure76.dta saved *create data for female fellow use figure76 gen fem =1 gen fel =1 save ff file ff.dta saved *create data for female nonfellow use figure76 gen fem =1 gen fel =0 save fnf file fnf.dta saved *create data for male fellow use figure76 gen fem =0 gen fel =1 save mf file mf.dta saved *create data for male nonfellow use figure76 gen fem =0 gen fel =0 save mnf file mnf.dta saved
Step 2.
use https://stats.idre.ucla.edu/stat/stata/examples/long/tobjob2, clear
(Academic Biochemists / S Long)
quietly tobit jobcen fem phd ment fel art cit, ll(1)
use ff, clear
predict ffcen, pr(.,1)
label var ffcen "female fellow"
sort phd
save ff, replace
file ff.dta saved
use fnf, clear
predict fnfcen, pr(.,1)
label var fnfcen "female nonfellow"
sort phd
save fnf,replace
file fnf.dta saved
use mf, clear
predict mfcen, pr(.,1)
label var mfcen "male fellow"
sort phd
save mf,replace
file mf.dta saved
use mnf, clear
predict mnfcen, pr(.,1)
label var mnfcen "male nonfellow"
sort phd
save mnf,replace
file mnf.dta saved
use ff, clear
merge phd using fnf mf mnf
graph twoway (scatter ffcen fnfcen mfcen mnfcen phd, c(l l l l) ///
xtitle("Ph.D. Prestige") ytitle("Pr(Censored)"))
Table 7.1, page 215. Hausman and Wise’s OLS and ML Estimates From a Sample With Truncation.
NOTE: This has been skipped because we do not have the data.