Benjamin Williams, "Controlling for Ability Using Test Scores", Journal of Applied Econometrics, Vol. 34, No. 4, 2019, pp. 547-565. The replication files contain the following: (1) Supplementary appendix (tsregressions-101218-wtable.pdf) (2) Replication files for the Monte Carlo exercises. All replication files are zipped in the file bw-files.zip. Except for the .dta file, they are ASCII files in DOS format. - montecarlos_all.m This is the file that simulates the model and calls various auxiliary files to obtain the estimates summarized in Tables 1 and B.1 in the paper. -that2pl.m This is a file that calculates the MLE estimates of the IRT scores according to the 2 parameter logistic (2PL) model. - kreg2.m This is a file that calculates a kernel regression estimator with leave-one-out cross validation bandwidth selection. - functionS.m This is a file containing the function used as the kernel by kreg2.m (3) Files for the NLSY application. The data used in the article are from the National Longitudinal Survey of Youth 1979, which is publicly available. The data was downloaded in two steps (on February 10 and February 21 of 2014) using the dbgator tool (https://www.nlsinfo.org/investigator/pages/login.jsp). The file nlsy_clean.do was used to clean the downloaded and save it as nlsy_v1_full.dta. The file nlsy_extract.do was then used to extract the data used for the analysis from nlsy_v1_full.dta. The replication files contain the following extracts but not nlsy_v1_full.dta. nlsy_v1_age30.dta, nlsy_v1_age30.csv -- These two files both contain all variables for all observations from nlsy_v1_full.dta with age=30. In .csv and .dta format. nlsy_matlab_chs_nwf.txt -- These files contain the extracts from nlsy_v1_full.dta that were used for analysis in Matlab (by the PLR.m nlsy_matlab_chs_nwm.txt code). The chs and hgc in the filename indicate which education variable is used. The nwf,nwm,wf,wm indicate . different subsamples (non-white female, non-white male, white female, white male, respectively). . . The Stata .do file summary_statistics.do was used to create Table 2. The Stata .do file regressions.do was used to create columns (1)-(3) in Table 3 and Figures 1 and 2. The Matlab .m file PLR.m (and auxiliary files kreg2.m and functionS.m) was used to estimate the partially linear regression model reported in column (4) in Table 3. Please address any questions to Ben Williams. bdwilliams [AT] gwu.edu. Ben Williams Dept. of Economics George Washington University 2115 G St. NW Washington, D.C. 20052