Estimates the over-dispersion parameter (theta) for a beta-binomial model stratified by sequencing depth, using panel of normal (PoN) BAF data.
Usage
EstimateTheta(normals_dt, pon_ref, n_bins)
Arguments
- normals_dt
Data frame or tibble. Panel of normal data.
- pon_ref
Data frame or tibble. Reference PoN data.
- n_bins
Integer. Number of depth strata to use.
Value
Returns a list with elements:
- theta_table
Data frame with columns depth_bin and theta (per-stratum estimates).
- breaks
Numeric vector of depth quantile breakpoints.
Details
The function compares observed BAF variance to expected binomial variance across bins and depths, and estimates theta as:
$$\theta = \frac{\mathrm{Var}(p)}{\mathrm{E}[p(1-p)/d]} - 1$$
where \(p\) is the observed BAF, and \(d\) is the median depth.
