Lab Notebook

What are the aims of the project? The question is stated from a modeling viewpoint.

Notation

$S_{ig}$: the $S$ statistic for individual (tissue sample) $i$ and gene $g$

Aims

1. dissect technical variation/noise?
   • yes: hierarchical model (parameters above $S_{ig}$)
   • no: use $S_{ig}$
2. remove confounding effects (of institution, RNA quality, population structure)?
   • yes: multiple regression of $S_{ig}$ for every gene, not only the known imprinted ones (fig S3)
   • no: regression is not necessary
3. allow unexplained variation across individuals within each gene?
   • yes: add latent variables and/or use hierarchical model
   • no: use only observed explanatory variables
4. variation across genes in the effect of age and other variables?
   • yes: hierarchical model
   • no: only regression
5. number/genome-wide extent of monoallelic expression interesting?
   • yes: hierarchical model with parameter $\pi_0$
   • no: presumed mono and biallelically expressing genes analyzed separately
6. $S_{ig}$ is a biased estimator of the expected fraction of maternal/paternal transcripts; is bias a concern?
   • yes: directly model read counts at each heterozyg SNP
   • no: adhere to $S_{ig}$

References

ms text

ms figures