Patches and recent notes

The design matrix may now exceed 2100 million non-zero cells, to facilitate fitting larger marker models..

Major improvement in time required to form a Large H inverse(201150 in pedigree, 40000 in marker based GRM).This was taking 9 hours and now takes 10 minutes.Using this H inverse takes 24 minutes per iteration. Fixed recent bug which prevented some large jobs from working(‘Insertion […]

Mainly bug fixes related to fitting large models, GLM and GRM models.

Allow access to 499 Gbyte (but no very large jobs have been tested yet)Better accommodation of long factor labels (ID !A a !LL 25 !L Gid.txt) ginv.sgiv !KEEP !SAVE reinverts the G inverse to obtain the G inverse, write it to ginv.sgrmand make it available […]

I have updated mPQL, an alternative to PQL which works better than regular PQL for ‘animal’ models when the average incidence of a binary trait is low. I am writing a paper describing the method and you are welcome to ask me for a preview. […]

Revised !SLOW qualifier actionRevised GLMM VC updatingAdded advice when model fails to convergeAdded some on screen guidance. Back transform predictions from GLM modelsUpdate processing of MARGINAL GLMM models (!M in model line flags MARGINAL random terms)Update Variogram graphics figurePLOT X (i.e. no Y variable specified) […]

Back transform predictions from GLM modelsUpdate processing of MARGINAL GLMM models (!M in model line flags MARGINAL random terms)Update Variogram graphics figure

Sorry: Internally it is labelled 2.08 Apr 16 2026 Updated !BINOMIAL !MARGINAL to expect !M rather than !R as the ‘random’ flag when fitting a binomial model with some random effects terms fitted marginally. i.e. usewk16 !bin !MARG ~ mu spec prov !m fam*reprather thanwk16 […]

Added !SELECT qualifier to !MBF lineFixed stack overflow bug in E70Cinverse Fixed bug in transformations on missing data.

Implemented some cosmetic changes suggest by Mario includingrelabelling the Fitted values in the .esy file PValues ==> Fitted_Values