Is there an equivalent to Stata's PSACALC in R?

user_10560648 0
Viewed 216

In a class, we have been utilizing PSACALC in Stata for Oster's bounding for understanding the degree of omitted variable bias in regression. I'm wondering if there is an alternative anyone is aware of in R.

stata r
1 Answers

I adapted the code from PSACALC and made an R package about two years ago. I've never gotten around to putting it on CRAN (don't know how), but you can install it from my GitHub. Here's a short example reproducing Table 5 from Oster (2019)

#install.packages("devtools")
library(devtools)
#devtools::install_github("siverskog/oster")
library(oster)

data(nunn)

# FORMULA FOR INTERMEDIATE (TILDE) REGRESSION

f01c <- trust_relatives ~ ln_export_area + age + age2 + male + urban_dum +
  as.factor(education) + as.factor(occupation) + as.factor(religion) +
  as.factor(living_conditions) + district_ethnic_frac +
  frac_ethnicity_in_district + as.factor(isocode) + malaria_ecology +
  total_missions_area + explorer_contact + railway_contact + cities_1400_dum +
  as.factor(v30) + v33_alt + ln_init_pop_density

# FORMULA FOR SHORT (O) REGRESSION

f01u <- trust_relatives ~ ln_export_area + as.factor(isocode)

# FITS 

fit01c <- lm(f01c, data = nunn)

nunn$infit01c <- is.element(rownames(nunn), names(fit01c$residuals))
fit01u <- lm(f01u, data = nunn, subset = infit01c) # USE ONLY OBS. IN TILDE FIT

# CALCULATE BETA/DELTA/RMAX WITH oster() TO
# REPRODUCE ROW 1 OF TABLE 5 IN OSTER (2019)

z <- oster(fit01u, fit01c, "ln_export_area")
b13 <- oster(fit01u, fit01c, "ln_export_area", rm = 1.3)$beta

round(c(z$input$beta_o, z$input$beta_tilde, z$beta, b13, z$rmax), 3)
user_49032760
Related