Learn R Programming

metaSEM (version 2.0)

Cor2DataFrame: Convert correlation or covariance matrices into a dataframe of correlations or covariances with their sampling covariance matrices

Description

Variable and appended moderator column names must not contain periods; rename them consistently, e.g., x.1 to x_1, and check for duplicates. Multi-column moderator names generated with periods are converted to underscores; collisions are rejected. Study names may contain periods. Generated pair labels that would collide are made unique with a _dup suffix; existing objects with ambiguous duplicate labels must be regenerated.

It converts the correlation or covariance matrices into a dataframe of correlations or covariances with their asymptotic sampling covariance matrices. It uses the asyCov at the backend.

Usage

Cor2DataFrame(
  x,
  n,
  v.na.replace = TRUE,
  cor.analysis = TRUE,
  acov = c("weighted", "individual", "unweighted"),
  Means,
  row.names.unique = FALSE,
  append.vars = TRUE,
  asyCovOld = FALSE,
  ...
)

Value

A list of components: (1) a data frame of correlations or covariances with their sampling covariance matrices; (2) a vector of sample sizes; (3) labels of the correlations; and (4) labels of their sampling covariance matrices.

Arguments

x

A list of data with correlation/covariance matrix in x$data and sample sizes x$n. Additional variables in x, in any order relative to data and n, can be attached. Named matrices are aligned to the first matrix by variable name.

n

If x is a list of correlation matrices without x$data and x$n, a vector of sample sizes n must be provided.

v.na.replace

Logical. For weighted or unweighted average sampling covariance matrices, replace missing covariance entries by their available weighted or unweighted mean before calculating sampling covariances. An entry missing from every study cannot be replaced and raises an error.

cor.analysis

Logical. The output is either a correlation or covariance matrix.

acov

If it is weighted, the average correlation/covariance matrix is calculated based on the weighted mean with the sample sizes. The average correlation/covariance matrix is used to calculate the sampling variance-covariance matrices.

Means

An optional matrix of means. The number of rows must be the same as the length of n. The sampling covariance matrices of the means are calculated by the covariance matrices divided by the sample sizes. Therefore, it is important to make sure that covariance matrices (not correlation matrices) are used in x when Means are included; otherwise, the calculated sampling covariance matrices of the means are incorrect.

row.names.unique

Logical. If it is FALSE (the default), unique row names are not created. If TRUE, the rows are named by the names of the studies, made unique.

append.vars

Whether to append the additional variables to the output dataframe.

asyCovOld

Whether to use the legacy Hessian-based method. It requires complete covariances among retained variables in its working matrix. asyCov() is preferred. See asyCov.

...

Further arguments to be passed to asyCov.

Author

Mike W.-L. Cheung <[email protected]>

See Also

asyCov, osmasem, create.vechsR, create.Tau2, create.V

Examples

Run this code

# \donttest{
## Provide a list of correlation matrices and a vector of sample sizes as the inputs
my.df1 <- Cor2DataFrame(Nohe15A1$data, Nohe15A1$n)

## Add Lag time as a variable
my.df1$data <- data.frame(my.df1$data, Lag=Nohe15A1$Lag, check.names=FALSE)

## Data
my.df1$data

## Sample sizes
my.df1$n

## ylabels
my.df1$ylabels

## vlabels
my.df1$vlabels

#### Simplified version to do it
my.df2 <- Cor2DataFrame(Nohe15A1)
# }

Run the code above in your browser using DataLab