vdorie / npci Goto Github PK

Hi Vdorie,

I saw your confusion that everyone is using your implementation for IHDP dataset simulation. Haha, I thought it is because of this highly cited NeurIPS 2016 paper, the auther claimed that they used your npci repo. I know your repo also due to this paper.

Get back to my confusion, I found out that you responded to the simulation setting things "is the setting A, B and C same to Jennifer Hill's setting A, B and C" with Yes answer. But what I found from your R code (from data.R file under ihdp_sim folder) is like:

From my understanding, if we set setting=="A", for the highlighted area, your algorithm will go to the else part isn't it? But, the response surface A in Jennifer's paper is stated to be linear, and I also refer to her code as follows:

The response surface B, which is non-linear, is similar to your code under the else condition. And if we set setting=="B", it will go through the linear response function under the if condition, which supposed to be non-linear in Jennifer's paper.

Thus, I just wanna confirm that, do we mess up the setting A and B here? Or if I miss something?

Thanks for any reply in advance!

Regards,
Hechuan

Hey there,

first off, thanks for implementing this package that has found so wide usage for evaluations.
I was wondering whether the IHDP setting A in your code refers to the setting A proposed by Hill in her paper?

I find it hard to get a clear overview over all the different datasets based on IHDP used for comparison of causal effect estimators. This would make things a bit clearer.

Thanks!

Could you please clarify how to use the package? I would like to generate 1k files similar to those found in https://github.com/claudiashi57/dragonnet/tree/master/dat/ihdp/csv in order to compare my approach to Dragonnet.

I want get the different biased degree data.How can get it?

Hi, there!

First, thanks for sharing the package for us!
I would have this package to have several simulations for IHDP. But when I was trying to install this package, it gave me the error message below:

* installing *source* package ‘npci’ ...
** using staged installation
** libs
clang -dynamiclib -Wl,-headerpad_max_install_names -undefined dynamic_lookup -single_module -multiply_defined suppress -L/Library/Frameworks/R.framework/Resources/lib -L/usr/local/lib -o npci.so npci.o -L/Library/Frameworks/R.framework/Resources/lib -lRblas -L/usr/local/gfortran/lib/gcc/x86_64-apple-darwin18/8.2.0 -L/usr/local/gfortran/lib -lgfortran -lquadmath -lm -F/Library/Frameworks/R.framework/.. -framework R -Wl,-framework -Wl,CoreFoundation
ld: warning: directory not found for option '-L/usr/local/gfortran/lib/gcc/x86_64-apple-darwin18/8.2.0'
ld: warning: directory not found for option '-L/usr/local/gfortran/lib'
ld: library not found for -lgfortran
clang: error: linker command failed with exit code 1 (use -v to see invocation)
make: *** [npci.so] Error 1
ERROR: compilation failed for package ‘npci’
* removing ‘/Library/Frameworks/R.framework/Versions/4.1/Resources/library/npci’

I have tried many ways like searching the problem, changing R version to 3.1.1, and even switching from Windows to MacOS to solve this issue but all failed.
Would you have any suggestion for this?

Thanks a lot!

Hello, sorry to bother you. Could you please tell me where the file 'examples/ihdp_sim/loadData.R' is? I couldn't find it in the repository. It seems that the data generation speed is a bit slow when using 'examples/ihdp_sim/runLocally.R' to generate data.So I need 'loadData.R.' If it's convenient, could you please let me know how to obtain it? I would greatly appreciate it.