28 March 2014


Next post Previous post


Changed to using meld for merging conflicted files in git. AMAZING!

Ecolunch - KM talk

Gotelli lab meeting - AD practice talk


Continuing from yesterday, spent much of day trying to use lmrob function for robust regression. Constant string of problems. Discussed with NG and one issue is that methods are best applied to large data sets, so doesn’t really make sense to use with only 22 points. Advantage of OLS is that it can nearly always fit a model. Simpler, and often robust, approach is to simply fit model with log transformed response variable. Use log(TPM+1) to deal with 0 values.

Power of git - checkout out old branch, removing all the lmrob code, and re-committed as HEAD of my master branch.

In the process of working on this, learned some useful tips for plyr:

  • add progress bar to see status of function

    .progress = “text”

  • get information for debugging on which piece of dataset problems are occuring using .inform from stackoverflow

    .inform = TRUE

Creative Commons Licence
This work is licensed under a Creative Commons Attribution 4.0 International License.