Maintenance log

[cwolock]

This Discussion reflects the current status of all major functions in rigr. Specific bugs/enhancements live in Issues; this post is meant to be a general tracking tool.

[cwolock]

descrip

[taylorokonek]

Basic unit tests written and additional error handling added. Minor changes to output for non-Surv variables and removal of version parameter. Documentation updated as well. All helper functions located in descrip_utils.R, and everything is documented well. Overall, this function is good to go barring any bugs we find in the future or additional enhancements we want.

[cwolock]

regress

[taylorokonek]

All helper functions for regress moved into regress_utils.R. Unit tests in progress. A few bugs and enhancements remain (#39 #12 #11). Following discussion #36, we will be removing gee and survival capabilities from the function to make it more maintainable moving forward (in progress). Further documentation needs to be done in regress_utils.R (#38), and many of the helper functions could likely be replaced with stringr functions.

[taylorokonek]

Many more unit tests written (coverage is about 75% at the moment). The main regress function is documented line by line in the code now. All of the survival capabilities should be removed at this point, some gee code chunks still remain but will be deleted soon. A lot of bugs and coefficient labeling in output has been fixed, just a few use cases left to address (see #68, for example). Have yet to look into lspline, dummy, and polynomial use cases in U.

[taylorokonek]

lspline functionality has been removed, dummy functionality has been debugged, and more unit tests have been written. All gee capabilities should now be removed. Most bugs have been addressed. Looking into polynomial usage, enhancements, and further unit testing remains.

[taylorokonek]

polynomial functionality is working, intercepts can now be removed from models with a "-1" in the formula or specifying intercept = FALSE, and multi-way (three-way or more) interactions are now functional

[cwolock]

wilcoxon

[cwolock]

Basic unit tests written and major (known) bugs fixed. Could use additional unit tests for output not provided by stats package's wilcox.test function, but as of now the comparable pieces match wilcox.test.

[cwolock]

ttest

[cwolock]

Unit tests for error handling written. Substantial bugs remain (see #42 #41 #40), although so far they do not affect inference.

[cwolock]

Unit tests for return values written. This includes one-sample, two-sample matched, two-sample unmatched, assuming equal and unequal variances, etc. Also streamlined the code a bit - eliminated the strat parameter which was identical to by. If we decide to split proportion tests off from ttest, then unit tests will be essentially done for this function.

[cwolock]

ttesti

[cwolock]

I reworked the code for this so that ttesti returns an object of class ttesti which then has a print method. I also harmonized the printed output with ttest.

[cwolock]

Unit tests for error handling and correct output are finished.

[taylorokonek]

bplot

[taylorokonek]

Currently, work on bplot has been put on hold since other functions (such as regress, ttest, etc.) are higher priority. We've discussed converting bplot so that it produces ggplot objects rather than base R plots, and I have a drafted version of the function on my personal machine. The function will have the same capabilities as the current bplot function, and users will not need to know ggplot in order to use it (it will remain a wrapper). We've also discussed changing some of the default parameters so that the plots are less crowded by default.

[cwolock]

lincom

[cwolock]

I have made some minor changes to lincom; in general, the function seems to be in good working order. I rewrote some of the code so that lincom returns an object versus just printing output. I also removed the overall test (although I can easily put it back in - see #62). Printing capabilities have been moved to lincom_utils to be consistent with other functions. Next task is unit testing.

[cwolock]

Unit tests for error handling and basic functionality have been written. I fixed a few minor bugs in the printed output.

[cwolock]

proptest

[cwolock]

This is a new function building off of the ability of old ttest to do tests of proportions. At the moment, proptest can do a one-sample proportion test (either approximate or exact, based on the binomial probabilities) or a two-sample test comparing two proportions (approximate only). Unit tests to be added.

[cwolock]

Unit tests have been added and code coverage is 100%. There are some enhancements that could be made - see #58.

[cwolock]

proptesti

[cwolock]

Same functionality as proptest, with with counts of successes and trials as opposed to binary data. Unit tests to be added.

[cwolock]

Unit tests have been added and code coverage is 100%. There are some enhancements that could be made - see #58.