Chris Hammill
Chris Hammill
If the LHS is set in `anatLm` the code breaks in mysterious ways. I should add a check for this. Two sided formulae are nonsensical here anyway.
Antoine ran into an interesting problem where he was trying to do `hanatLmer` on volume proportions. The model failed because at the root node there is zero variance (every brain...
Zsu ran a `mincLm` where the t-statistics for one column were shifted relative to all the others, in `mincPlotSliceSeries` this appeared as the brain showing up on the bottom edge...
The bounds for the mincLm randomization test are perhaps too tight and fail on some runs. This is an irritating test without ground truth, maybe a better test is needed...?
This is a change in behaviour at some point in the history of `testthat`, tests are run with the working directory set to the location of the test file instead...
Dealing with `*Lm` and `*Lmer` objects is a nightmare for doing anything non-standard. The big issue is attribute tracking and the imperfect mapping of `stat-type` and `df` attributes to columns....
Apparently when the background anatomy is not of the usual scale e.g. FA in [0,1] the locator plot isn't useful. Brought to my attention by Jacob.
Brought to my attention by Jacob.
For future matching adventures