noiseBgEffects simulates observational noise with a proportion of the effect shared across samples and a proportion independent across samples.
noiseBgEffects(N, P, mean = 0, sd = 1, sampleID = "ID_", phenoID = "Trait_", shared = TRUE, independent = TRUE, id_samples = NULL, id_phenos = NULL)
| N | Number [integer] of samples to simulate. |
|---|---|
| P | Number [integer] of phenotypes to simulate. |
| mean | Mean [double] of the normal distribution. |
| sd | Standard deviation [double] of the normal distribution. |
| sampleID | Prefix [string] for naming samples. |
| phenoID | Prefix [string] for naming traits. |
| shared | [bool] shared effect simulated if set to TRUE; at least one of shared or independent has to be set to TRUE. |
| independent | [bool] independent effect simulated if set to TRUE. |
| id_samples | Vector of [NrSamples] sample IDs [string]; if not provided constructed by paste(sampleID, 1:N, sep=""). |
| id_phenos | Vector of [NrTraits] phenotype IDs [string]; if not provided constructed by paste(phenoID, 1:P, sep=""). |
Named list of shared noise effects (shared: [N x P] matrix) and independent noise effects (independent: [N x P] matrix), the covariance term of the shared effect (cov_shared: [P x P] matrix) and the covariance term of the independent effect (cov_independent: [P x P] matrix).
For the simulation of the observational noise effects, two components are used: i) matrix B [N x P] with vec(B) drawn from a normal distribution with mean=mean and sd=sd and ii) the trait design matrix A [P x P]. For the independent effect, A is a diagonal matrix with normally distributed values. A for the shared effect is a matrix of rowrank one, with normally distributed entries in row 1 and zeros elsewhere. To construct the final effects, the two matrices are multiplied as: E = BA^T.
noiseBG <- noiseBgEffects(N=100, P=20, mean=2)