Compute log fractions and precision weights from matrix of c ounts, where columns are variables and rows are samples
Value
An EList object with the following components:
- E:
numeric matrix of log transformed counts
- weights:
numeric matrix of observation-level inverse-variance weights
Details
For real data, the asymptotic variance formula can give weights that vary substantially across samples and give very high weights for a subset of samples. In order to address this, we regularize the weights to reduce the variation in the weights to have a maximum ratio of max.ratio between the maximum and quant quantile value.
Examples
# set probability of each category
prob <- c(0.1, 0.2, 0.3, 0.5)
# number of total counts
countsTotal <- 300
# number of samples
n_samples <- 100
# simulate info for each sample
info <- data.frame(Age = rgamma(n_samples, 50, 1))
rownames(info) <- paste0("sample_", 1:n_samples)
# simulate counts from multinomial
counts <- t(rmultinom(n_samples, size = countsTotal, prob = prob))
colnames(counts) <- paste0("cat_", 1:length(prob))
rownames(counts) <- paste0("sample_", 1:n_samples)
# run logFrac on counts
cobj <- logFrac(counts)
# run standard variancePartition analysis on crumblr results
library(variancePartition)
fit <- dream(cobj, ~ Age, info)
fit <- eBayes(fit)
topTable(fit, coef = "Age", sort.by = "none")
#> logFC AveExpr t P.Value adj.P.Val B
#> cat_1 0.0003227888 -2.4019876 0.09838761 0.9218201 0.9218201 -8.863774
#> cat_2 -0.0012459745 -1.6944100 -0.64903768 0.5177915 0.9218201 -8.599812
#> cat_3 0.0009504719 -1.3192849 0.60346707 0.5475549 0.9218201 -8.625023
#> cat_4 -0.0001702998 -0.7906898 -0.16652797 0.8680754 0.9218201 -8.785482