All estimands in AlphaPeel are in one way or another connected to genetic causes and their parameters. However, sometimes observed genotypes are impacted by non-genetic effects. This might sound strange (env should not impact DNA!), but who is genotyped can be impacted by non-genetic effects and not accounting for this in peeling can lead to wrong estimates of founder allele probabilities etc.
So, we should think how could we expand the AlphaPeel's model with user defined effects (factors with levels) and associated conditional probability tables that would impact the observed genotype. This would come in as an effect on the observed genotype (so it impacts the data we see/collect) but as independent of all other variables in the model (as it should). When many observed genotypes are connected to such a variable we would then be able to estimate its effect on the data we see.
In a way we are building GLM like functionality here ...
This should probably be done after penetrance work #74
All estimands in AlphaPeel are in one way or another connected to genetic causes and their parameters. However, sometimes observed genotypes are impacted by non-genetic effects. This might sound strange (env should not impact DNA!), but who is genotyped can be impacted by non-genetic effects and not accounting for this in peeling can lead to wrong estimates of founder allele probabilities etc.
So, we should think how could we expand the AlphaPeel's model with user defined effects (factors with levels) and associated conditional probability tables that would impact the observed genotype. This would come in as an effect on the observed genotype (so it impacts the data we see/collect) but as independent of all other variables in the model (as it should). When many observed genotypes are connected to such a variable we would then be able to estimate its effect on the data we see.
In a way we are building GLM like functionality here ...
This should probably be done after penetrance work #74