Summary Method for Multiple Imputation Linear Models

DESCRIPTION:

Returns a summary list for a linear model, which consolidates information across multiple imputations.

USAGE:

miSummary(miList containing lm objects) 
miSummary.lm(object1, object, correlation = T) 

REQUIRED ARGUMENTS:

object1
An lm object, which should be the same as the first imputation from object.
object
a multiple imputation object whose completed data sets (imputations) are lm objects.

OPTIONAL ARGUMENTS:

correlation
logical flag: if TRUE, then the correlation matrix for the coefficients is included in the summary.

VALUE:

a list is return with components:
call, terms
the call and terms components of the lm objects.
residuals
the model residuals from the lm objects.
coefficients
a matrix with five columns, containing the coefficients, standard errors, t-values, degrees of freedom, and p-values.
sigma
the residual standard error estimate.
df.orig
degrees of freedom from each complete data analysis; number of degrees of freedom for the model and for residuals.
df.coef
estimated degrees of freedom for each regression coefficient; this is smaller than the residual degrees of freedom in df.orig because of additional uncertainty due to missing values.
r.squared
multiple R-square statistic.
fstatistic
numeric vector of length three giving the F test for the regression. The first element is the statistic and the last two elements are the degrees of freedom.
cov.unscaled
the unscaled covariance matrix; i.e, a matrix such that multiplying it by sigma^2 produces an estimated covariance matrix for the coefficients.
correlation
the computed correlation coefficient matrix for the coefficients in the model.

DETAILS:

Some of the components above are miList objects, containing results for each imputation. Others are consolidated results, combining information from all imputations. In particular,
sigma is the square root of the average squared value of sigma from the multiple imputations,
r.squared is computed as the ratio of averages of numerator and denominator sums of squares across imputations,
fstatistic,
df.coef, and
cov.unscaled are obtained using consolidation methods described in the references below.

REFERENCES:

Hesterberg, T. (1998), Combining multiple imputation t, chi-square, and F inferences , Insightful Technical Report number 75.

Rubin, D. B. (1987), Multiple imputation for nonresponse in surveys , John Wiley, New York.

Schafer, J. L. (1997), Analysis of Incomplete Multivariate Data , Chapman & Hall, London.

SEE ALSO:

, , , .

EXAMPLES:

fit <- miEval(lm(chol14~., data = cholesterolImpExample)) 
miSummary(fit)