Hi,
I’ve created a simple fixed effects version of Li and Racine’s Kernel
Regression with Mixed Data Types (npreg function from np package) by
including an unordered firm id variable and ordered year variable as extra
covariates. See below for the setup:
******
bandwidth <- npregbw(formula = log.PPE ~ factor(tax.policy) + tax.rate +
GDP.growth + log.firm.turnover + MNE.country.count, factor(subsidiary_id),
ordered(year)
data = Norway,
regtype = "ll",
bwmethod = "cv.aic",
ckertype = "gaussian",
ckerorder = 2,
ukertype = "aitchisonaitken",
okertype = "wangvanryzin" )
model.np <- npreg(bws = bandwidth, gradients = TRUE, residuals = TRUE)
plot(model.np,
plot.errors.method = "bootstrap",
plot.errors.boot.num = 25,
plot.errors.boot.method = "inid")
******
I understand that nonparametric regressions allow for nonlinear
relationships for all covariates and interactions between all covariates. I
have calculated the gradients in the npreg function but would like advice
about how to use them.
I have several related questions:
1. How do I calculate the average partial effect of the binary tax policy
variable over all subsidiaries in the sample? That is, I would like to
calculate the:
a. average investment level of each subsidiary when the policy variable
= 0 (holding other covariates at their median for that subsidiary) and
b. average investment level of each subsidiary when the policy variable
= 1 (holding other covariates at their median for that subsidiary) and then
c. take the difference.
d. And then I would like to calculate the average of this difference
over all subsidiaries.
2. Similarly, I am interested in the interactions between the policy
variable and other covariates. For example, was the policy effect bigger
for larger subsidiaries or for multinationals with a presence in many other
countries?
Basically, I would like to know how to produce the type of output that one
can get from the 'margins' command in STATA.
3. Finally, is there a way to customise the partial regression plots that
result from the plot command after using npreg? Eg. How do I choose to
hold the policy variable at 0 and at 1 so that I can compare two versions of
the plots?
Many thanks,
Lisa
Masters Student,
University of Tübingen.
--
View this message in context:
http://r.789695.n4.nabble.com/npreg-average-partial-effects-and-customised-partial-regression-plots-tp4710500.html
Sent from the R help mailing list archive at Nabble.com.
______________________________________________
R-help@r-project.org mailing list -- To UNSUBSCRIBE and more, see
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
