Update to -psacalc- on SSC

[Jorge Eduardo Perez Perez]

On behalf of Emily Oster, and thanks to Kit Baum as usual, an update package to -psacalc- has been uploaded to SSC.

The updated command has been rewritten in Mata and now it allows factor variables, if/in qualifiers and linear panel data models.

psacalc: module to calculate treatment effects and relative degree of selection under proportional selection of observables and unobservables.

psacalc is performed after linear models to evaluate the possible degree of omitted variable bias under the assumption that the selection on the observed controls is proportional to the selection on the unobserved controls. Details of the calculation and theory are in Oster (2016), JBES, Forthcoming. The command requires the user to identify an independent variable of interest from the regression. The command calculates (1) a value of the relative degree of selection to match a given treatment effect or (2) a value of the treatment effect given a relative degree of selection. Options are given to specify a maximum R-squared in absence of measurement error and a set of observable controls not informative about the observables.

-psacalc- uses Mata and requires Stata 12

To update type;

adoupdate psacalc, update

To install for the first time, type;

ssc install psacalc.

[Leon Schmidt]

Thanks a lot for providing this program!

One question: What is the correct way to calculate Oster bounds for interaction terms in a panel data model with fixed effects? Would the following (nonsensical) example be correct?

Thanks a lot for your help!

webuse nlswork, clear
xtset idcode year
xtreg ln_wage c.ttl_exp##i.race i.year, fe
psacalc delta ttl_exp, mcontrol(2.race#c.ttl_exp 3.race#c.ttl_exp)
psacalc delta 2.race#c.ttl_exp, mcontrol(ttl_exp 3.race#c.ttl_exp)

[Leon Schmidt]

I would just like to push my question above again in case someone knows how to properly code Oster bounds for interactions. Sorry for tagging you, but maybe you Jorge Eduardo Perez Perez know? Thank you very much!