inst/extdata/pipeline.md

In VanAndelInstitute/bifurcatoR: Hunt for bumps in high-dimensional (or low-dimensional!) data

Basic documentation of current (i.e. as of 2/14/22) simulation framework. To be updated as code develops/is modified.

Generate Data

configure_params

Based on specific parameters (see below), create an array of simulated data to be used fitted on various methods. Alternatively, we can generate the data as it is being fitted to a method of interest but I figured it might just be easier to simulate the data beforehand, especially if we are interested in multiple configurations.

Right now, we are only interested in the Normal distribution, but configure_params has been set up to include: Beta Cauchy Normal Uniform * Weibull

The other parameters include: Balanced Design (balanced): whether to sample X=0 and X=1 (for strain effect) with equal probability from the Binomial distribution. Number of Groups (nGroups): n_groups = 2 for strain effect; n_groups = 3 for allele effect (TBD). * Number of Subjects (nSubjects): self-explanatory.

The output of configure_params is a list of arguments that correspond to the function of interest, e.g. rnorm takes in the mean parameter whereas rweibull takes in the scale parameter.

X_sim

When generating the X variable via X_sim, balanced samples in X are obtained using sample to get exactly a 50:50 split but unbalanced design can either take in a prob argument such that we have rbinom(nSubjects, 1, prob) or default to a 25:75 split (hard-coded within the function).

y_sim

Given the output from X_sim, generate data that correspond to the null, mean-effect, variance-effect, and mean-variance-effect models.

• Null Model: rnorm(nSubjects, mean = 1, sd = 0) (default parameters)
• Mean Model: X * mean_effect_size + rnorm(nSubjects)
• Variance Model: For X=0, rnorm(nSubjects); for X=1, rnorm(nSubjects, sd = variance_effect_size)
• Mean-Variance Model: For X=0, rnorm(nSubjects, sd = variance_effect_size); for X=1, mean_effect_size + rnorm(nSubjects, sd = variance_effect_size)

Additionally, the data above were normalized by the Median Absolute Deviation (MAD):

VanAndelInstitute/bifurcatoR documentation built on Oct. 13, 2024, 6:29 p.m.