- Directory FNS contains the main functions for the simulations and data applications. Specifically,
- FNS_DataGene_Simu1 : function to generate data for simulation 1
- FNS_DataGene_Simu2 : function to generate data for simulation 2
- FNS_SmoBack : functions to conduct smoothbackfitting
- FNS_SmoBack_credit : differs from FNS_SmoBack only in the stopping criterion
- Directory datasets contains the datasets credit and flight for data applications. Due to the upload restrictions, we provide the preprocessed data here, which are seperated into several parts. Please merge them before processing. Raw data are available in the following links:
- credit : https://data.mendeley.com/datasets/wb3ndt69gf (P2P_Macro_Data.dta, 3 GB)
- flight : https://community.amstat.org/jointscsg-section/dataexpo/dataexpo2009
-
Directory res contains the analysis results of the simulations and data applications.
-
Directory fig contains the figures generated by the numerical experiments.
-
Directory Compare_Code contains the code for simulation 3 to compare with other online nonparametric methods in the degenerated one-dimensional case.
The R scripts simulation1.R and simulation2.R contain the code for the simulation study which computes the integrated mean square errors of the online and batch estimates as well as computing time, respectively.
The R scripts simu_figplot.R and simulation2_figplot.R contain the code to generate Figure 4-7.
As the bandwidth selection is not our focus, all procedures of estimating the constants for bandwidths in the simulations and data applications are collected in the R script supp-bandwidth_selection.R. The corresponding results are stored in the directory res. One can load them directly when reproducing the numerical experiments.
The R script Compare_Code/simuscript.R contains the code to compute the integrated mean square errors of different online nonparametric methods in the degenerated one-dimensional case.
The R script Compare_Code/figplot.R contains the code to generate Figure 1 of the supplement.
The R scripts flight_preprocess.R contains the code to preprocess the raw csv data which can be downloaded from the above link. The preprocessed data are given in the Directory datasets.
The R scripts flight_main.R and credit_main.R contain the code which computes the online and batch estimates as well as computing time for the two datasets, respectively.
The R scripts flight_figplot.R contains the code to generate table 1 and Figure 8 and credit_figplot.R contains the code to generate table 1 and Figure 8 contains the code to generate table 2 and Figure 9.