biocore / scikit-bio / 3660

0%

push

travis-ci

Permdisp (#1532)

* hit roadblock with permdisp, aggregate wont handle mixed type grouping

* tests for permdisp centroids works

* Main function now returns a series from build_results, same as permanova.
Pvalue now computed using monte_carlo method.
Started testing against the qiime script and so far the results do not match.
This may be due to how betadisper in R is using spatial medians, or the way we
are computing the centroids and F value is all wrong

* also added null matrix test

* added error raise tests

* I think I got medians to work

* this seems to work with hdmedians

* think we've worked out all the bugs, it is still slow, still havent decided on median implem

* refactoring and comments

* some refactoring

* totally flake8ed

* flake8ing

* marked non-explicitly tests functions

* problems with sphinx gwuh

* doing some refactoring for speed

* tried using loops. loops are bad

* weird commit issues. not sure what im doing

* init and test suite

* may have finally found right architexture for optimization

* cleaned up code, still need to rework some tests in test suite

* added hdmedians to setup.py. Do the benchmark comparisons to betadisper in R and submit a PR

* benchmarked R, gonna submit a pull request baby woohoo

* adds permdisp to skbio.stats.distance, fixes issue-1228

* should install hdmedians now

* fixed some tests that were failing Travis due to floating point rounding error

* trying again to fix travis rounding failures. last commit missed a few things

* one more fix for travis tests

* still trying to have travis work

* Update CHANGELOG.md

* just wanted to commit these small changes, not finished yet

* addressing pr comments

* I did not make flake8 adjustments in the last commit

* made jupyter notebook to compare permdisp results and look at the results of random matrices

* removing unneeded files

* addressed latest pr comments

* a... (continued)

15944 of 16380 branches covered (97.34%)

37736 of 38022 relevant lines covered (99.25%)

0.99 hits per line