baldur (version 0.0.4)

estimate_gamma_hyperparameters: Estimate Gamma hyperparameters for sigma

Description

Estimates the hyperparameters for the Bayesian data and decision model. estimate_gamma_hyperparameters is a wrapper that adds new columns to the data (one for alpha and one for betas). Note that for lgmr objects, the estimate_beta function assumes that the data is ordered as when the model was fitted. If this is not the case, theta's will be incorrectly matched with peptides---resulting in wrong estimates of beta parameters. On the other hand, estimate_gamma_hyperparameters will temporarily sort the data as when fitted and the sort it back as it was input to the function.

Usage

estimate_gamma_hyperparameters(reg, data, ...)
# S3 method for glm
estimate_gamma_hyperparameters(reg, data, ...)
# S3 method for lgmr
estimate_gamma_hyperparameters(reg, data, id_col, ...)
estimate_beta(reg, mean, ...)
# S3 method for glm
estimate_beta(reg, mean, alpha, ...)
# S3 method for lgmr
estimate_beta(reg, mean, m, s, ...)

Value

estimate_gamma_hyperparameters returns a tibble or data.frame

with the alpha,beta hyperparameters estimates as new columns.

estimate_beta returns estimates of the beta parameter(s)

Arguments

reg

A glm Gamma regression or a lgmr object

data

A tibble or data.frame to add gamma priors to

...

Currently not in use

id_col

A character for the name of the column containing the name of the features in data (e.g., peptides, proteins, etc.)

mean

The mean value of the peptide

alpha

The alpha parameter of the peptide

m

The mean of the means

s

The sd of the means

Examples

# Setup model matrix
design <- model.matrix(~ 0 + factor(rep(1:2, each = 3)))
colnames(design) <- paste0("ng", c(50, 100))

# Normalize data
yeast_norm <- yeast %>%
    psrn("identifier") %>%
    # Get mean-variance trends
    calculate_mean_sd_trends(design)

# Fit gamma regression (could also have been a lgmr model)
gam_reg <- fit_gamma_regression(yeast_norm, sd ~ mean)

# Estimate priors
gam_reg %>%
    estimate_gamma_hyperparameters(yeast_norm)

# Can also explicitly estimate the beta parameters
# Note this is order sensitive.
estimate_beta(gam_reg, yeast_norm$mean, 1/summary(gam_reg)$dispersion)