Guillermo del Angel 6b37350bc0 Two hairy bugs in pool caller: a) Site error model wasn't counting errors in insertions correctly - Alleles passed in had padded ref byte, but event base in PileupElement doesn't have it. As a result, mismatch rate was grossly overestimated with insertions and we missed several calls we should have made. Integration test reflects changes. b) Adding a ref GL to the exact model is correct mathematically but AFResult wasn't filled properly. As a result, QUAL was junk in pure ref sites, and in all other sites the last ref GL introduced wasn't properly updating Pr(AF>0). c) Added integration test that covers -out_mode EMIT_ALL_CONFIDENT_SITES. Not fully sure if the math is 100% correct (for both diploid and generalized case) but at least now diploid and non-diploid cases behave similarly. md5 of this new test will fail since it's taking me a long time to run so I'll update from Bamboo output shortly		2012-09-14 13:13:22 -04:00
..
src/org/broadinstitute/sting	Two hairy bugs in pool caller: a) Site error model wasn't counting errors in insertions correctly - Alleles passed in had padded ref byte, but event base in PileupElement doesn't have it. As a result, mismatch rate was grossly overestimated with insertions and we missed several calls we should have made. Integration test reflects changes. b) Adding a ref GL to the exact model is correct mathematically but AFResult wasn't filled properly. As a result, QUAL was junk in pure ref sites, and in all other sites the last ref GL introduced wasn't properly updating Pr(AF>0). c) Added integration test that covers -out_mode EMIT_ALL_CONFIDENT_SITES. Not fully sure if the math is 100% correct (for both diploid and generalized case) but at least now diploid and non-diploid cases behave similarly. md5 of this new test will fail since it's taking me a long time to run so I'll update from Bamboo output shortly	2012-09-14 13:13:22 -04:00
test/org/broadinstitute/sting/gatk/walkers	Two hairy bugs in pool caller: a) Site error model wasn't counting errors in insertions correctly - Alleles passed in had padded ref byte, but event base in PileupElement doesn't have it. As a result, mismatch rate was grossly overestimated with insertions and we missed several calls we should have made. Integration test reflects changes. b) Adding a ref GL to the exact model is correct mathematically but AFResult wasn't filled properly. As a result, QUAL was junk in pure ref sites, and in all other sites the last ref GL introduced wasn't properly updating Pr(AF>0). c) Added integration test that covers -out_mode EMIT_ALL_CONFIDENT_SITES. Not fully sure if the math is 100% correct (for both diploid and generalized case) but at least now diploid and non-diploid cases behave similarly. md5 of this new test will fail since it's taking me a long time to run so I'll update from Bamboo output shortly	2012-09-14 13:13:22 -04:00