random_effects function

Random effects structure

Use RE_cor to generate the random effects correlation matrix, under the assumption of common heterogeneity variance (i.e. all within-study correlations are 0.5). Use which_RE to return a vector of IDs for the RE deltas (0 means no RE delta on this arm).

RE_cor(study, trt, contrast, type = c("reftrt", "blshift"))

which_RE(study, trt, contrast, type = c("reftrt", "blshift"))

Arguments

• study: A vector of study IDs (integer, character, or factor)
• trt: A factor vector of treatment codes (or coercible as such), with first level indicating the reference treatment
• contrast: A logical vector, of the same length as study and trt, indicating whether the corresponding data are in contrast rather than arm format.
• type: Character string, whether to generate RE structure under the "reference treatment" parameterisation, or the "baseline shift" parameterisation.

Returns

For RE_cor(), a correlation matrix of dimension equal to the number of random effects deltas (excluding those that are set equal to zero).

For which_RE(), an integer vector of IDs indexing the rows and columns of the correlation matrix returned by RE_cor().

Examples

RE_cor(smoking$studyn, smoking$trtn, contrast = rep(FALSE, nrow(smoking)))
RE_cor(smoking$studyn, smoking$trtn, contrast = rep(FALSE, nrow(smoking)), type = "blshift")
which_RE(smoking$studyn, smoking$trtn, contrast = rep(FALSE, nrow(smoking)))
which_RE(smoking$studyn, smoking$trtn, contrast = rep(FALSE, nrow(smoking)), type = "blshift")