{smcl}
{* 14jul2004}{...}
{hline}
help for {hi:glmcorr}
{hline}
{title:Correlation measure of predictive power and RMS error for GLMs}
{p 8 17 2}{cmd:glmcorr} [ , {cmdab:jack:knife} ]
{title:Description}
{p 4 4 2}{cmd:glmcorr} calculates the correlation between the response and the
fitted or predicted response, its square, and the root mean square error after
{cmd:glm}.
{title:Remarks}
{p 4 4 2}Zheng and Agresti (2000) discuss the correlation between the
response and the fitted or predicted response as a general measure
of predictive power for GLMs. This measure has the advantages of
referring to the original scale of measurement, of
being applicable to all types of GLM and of being familiar to many
users of statistics. Preferably, it should be used
as a comparative measure for different models applied to the
same data set, given that restrictions on values of the response
may imply limitations on its value (see e.g. Cox and Wermuth, 1992).
{p 4 4 2}For an arbitrary GLM, this correlation is invariant under a
location-scale transformation and it is the positive square root of the average
proportion of variance explained by the predictors. However, again for an
arbitrary GLM, it need not equal the positive square root of other definitions
of R-square (e.g. Hardin and Hilbe, 2001); and it need not be monotone
increasing in the complexity of the predictors, although in practice that is
common. The correlation is necessarily sensitive to outliers.
{p 4 4 2}As the predicted is a function of the observed, the correlation
calculated from a sample may be expected to be biased upwards. A jackknifed
correlation is provided as one alternative. Zheng and Agresti provide more
discussion of this point, including other estimators and a bootstrap approach
to providing confidence intervals for the correlation and to estimating the
degree of overfitting.
{p 4 4 2}The root mean square error is calculated as the square root of the sum
of squares of (observed - fitted) divided by the residual degrees of freedom.
{title:Options}
{p 4 8 2}{cmd:jackknife} specifies that a jackknifed estimate of
the correlation be provided.
{title:Examples}
{p 4 8 2}{inp:. glm {it:whatever}}{p_end}
{p 4 8 2}{inp:. glmcorr}
{title:Author}
{p 4 4 2}Nicholas J. Cox, University of Durham, U.K.{break}
n.j.cox@durham.ac.uk
{title:References}
{p 4 4 2}Cox, D.R. and N. Wermuth. 1992. A comment on the coefficient of
determination for binary responses. {it:American Statistician} 46: 1-4.
{p 4 4 2}Hardin, J. and J. Hilbe. 2001.
{it:Generalized linear models and extensions.}
College Station, TX: Stata Press.
{p 4 4 2}Zheng, B. and A. Agresti. 2000. Summarizing the predictive power of a
generalized linear model. {it:Statistics in Medicine} 19: 1771-1781.
{title:Saved results}
{p 4 4 2}{cmd:r(rho)}{space 9}correlation between observed and predicted{p_end}
{p 4 4 2}{cmd:r(rsq)}{space 9}square of correlation between observed and predicted{p_end}
{p 4 4 2}{cmd:r(jrho)}{space 8}jackknifed correlation between observed and predicted (if requested){p_end}
{p 4 4 2}{cmd:r(rmse)}{space 8}root mean square error
{title:Also see}
{p 4 13 2}
Manual: {hi:[R] glm}, {hi:[R] jknife}{p_end}
{p 4 13 2}
On-line: help for {help glm}, {help jknife}