NEWS.md

In yuqimemeda/SortedEffects: Estimation and Inference Methods for Sorted Causal Effects and Classification Analysis

NEWS

Shuowen Chen 11/4/2019

SortedEffects 1.1.0.

This note document the changes from SortedEffects 1.0.0. The following changes are suggested by Norman Martloff, the R Journal editor, Thomas Leeper and an anonymous referee from the R Journal.

Major Changes

1. Compared to the previous version 1.0.0, the package now has a S3 structure. The plotting functions (plot.spe, plot.ca and plot.subpop) now are methods of generic plot().

2. In addition to the plotting methods, we provide three new summary methods (summary.spe, summary.ca and summary.subpop), which are methods of generic summary(). We believe these three new methods provide richer outputs to the users.

Minor Changes

1. We decaptalize the three functions (spe, ca, subpop).
2. We rename some arguments in the three functions to be snake case rather than dot case. These are var (instead of var.T), boot_type (instead of boot.type), range_cb (instead of range.cb), var_type (instead of var.type). We rename the argument B in the three functions to be b, and change the default to 500.
3. The arguments method, boot_type, var_type are have a finite set of options for users to choose from. So we now explicitly provide these options in the argument, and the code will use match.arg() to match user input for execution.
4. In each of the three functions (spe, ca, subpop) we add an option called parallel that allows users to either turn on or off the parallel computing. If users turn on the parallel computing, they can further specify how many CPUs to use via the cores option.
5. The package now has a progress bar showing the progress of bootstrap estimation. It also shows how many CPUs the users are using.
6. The option cl in function ca wasn't very intuitive. Now the users can choose from two alternatives (diff or hoth), which are the most common for cl.
7. In plot.subpop, we add an option called overlap that allow users to either keep or drop the overlapped observations in the plot.
8. In the previous version, the option t and cat in function ca required users to manually input all the variables. In particular, if some variables of interest are factor, the users need to manually create a lot of indicator variables and type them as an input. In the current version, users only need to input the factor variable itself, and the code will automatically generate indicator variables and run the estiamtions. We believe this makes the package easier to use.
9. In each function (spe, ca, subpop), we change the default of option taus to c(5:95)/100. 10 In function ca, we change the default to c(1:99)/100.

yuqimemeda/SortedEffects documentation built on Nov. 9, 2019, 3:49 a.m.