par.avg: Parameter averaging

Description

Average a coefficient with standard errors based on provided weights. This function is intended chiefly for internal use.

Usage

par.avg(x, se, weight, df = NULL, level = 1 - alpha, alpha = 0.05,
  revised.var = TRUE, adjusted = TRUE)

Value

par.avg returns a vector with named elements:

Coefficient: model coefficients
SE: unconditional standard error
Adjusted SE: adjusted standard error
Lower CI, Upper CI: unconditional confidence intervals.

Arguments

x: vector of parameters.
se: vector of standard errors.
weight: vector of weights.
df: optional vector of degrees of freedom.
alpha, level: significance level for calculating confidence intervals.
revised.var: logical, should the revised formula for standard errors be used? See ‘Details’.
adjusted: logical, should the inflated standard errors be calculated? See ‘Details’.

Author

Kamil Bartoń

Details

Unconditional standard errors are square root of the variance estimator, calculated either according to the original equation in Burnham and Anderson (2002, equation 4.7), or a newer, revised formula from Burnham and Anderson (2004, equation 4) (if revised.var = TRUE, this is the default). If adjusted = TRUE (the default) and degrees of freedom are given, the adjusted standard error estimator and confidence intervals with improved coverage are returned (see Burnham and Anderson 2002, section 4.3.3).

References

Burnham, K. P. and Anderson, D. R. 2002 Model selection and multimodel inference: a practical information-theoretic approach. 2nd ed.

Burnham, K. P. and Anderson, D. R. 2004 Multimodel inference - understanding AIC and BIC in model selection. Sociological Methods & Research 33, 261--304.