acml.lmem2: Fitting function: ACML or WL for a linear mixed effects model...

In schildjs/ods4lda: Outcome Dependent Sampling (ODS) Designs for Linear Mixed Models Data Analysis

Description

Fitting function: ACML or WL for a linear mixed effects model (random intercept and slope)

Usage

acml.lmem2(formula.fixed, formula.random, data, id, w.function = "mean",
  InitVals, cutpoints, SampProb, Weights, ProfileCol = NA)

Arguments

formula.fixed	formula for the fixed effects (of the form y~x)
formula.random	formula for the random effects (of the form ~z). Right now this model only fits random intercept and slope models.
data	data frame that should contain everything in formula.fixed, formula.random, id, and Weights. It does not include: w.function, cutpoints, SampProb
id	sum(n_i) vector of subject ids (a variable contained in data)
w.function	sum(n_i) vector with possible values that include "mean" (mean of response series), "intercept" (intercept of the regression of Yi ~ zi where zi is the design matrix for the random effects (solve(t.zi* zi) * t.zi)[1,]), "intercept1" (intercept of the regression of Yi ~ zi where zi is the design matrix for the random effects (solve(t.zi * zi) * t.zi)[1,]). "intercept2" (second intercept of the regression of the Yi ~
InitVals	starting values for c(beta, log(sigma0), log(sigma1), log((1+rho)/(1-rho)), log(sigmae))
cutpoints	A matrix with the first dimension equal to sum(n_i). These cutpoints define the sampling regions for individual subjects. If using a low, medium, high, sampling scheme, this is a sum(n_i) by 2 matrix that must be a distinct object not contained in the dat dataframe. Each row is a vector of length 2 c(k1,k2) to define the sampling regions, i.e., low, middle, high. If using a square doughnut design this should be sum(n_i) by 4 matrix (var1lower, var1upper, var2lower, var2upper). Each subject should have n_i rows of the same values.
SampProb	A matrix with the first dimension equal to sum(n_i). Sampling probabilities from within each region. For low medium high sampling, each row is a vector of length 3 with sampling probabilities for each region. For bivariate stratum sampling each row is a vector of length 2 with sampling probabilities for the inner and outer strata. Each subject should have n_i rows of the same values. Not in data.
Weights	Subject specific sampling weights. A vector of length sum(n_i). This should be a variable in the data dataframe. It should only be used if doing IPWL. Note if doing IPWL, only use robcov (robust variances) and not covar. If not doing IPWL, this must be a vectors of 1s.
ProfileCol	the column number(s) for which we want fixed at the value of param. Maximizing the log likelihood for all other parameters while fixing these columns at the values of InitVals[ProfileCol]

Value

Ascertainment corrected Maximum likelihood: Ests, covar, LogL, code, robcov

schildjs/ods4lda documentation built on March 16, 2020, 8:16 a.m.