------------------------------------------------------------------------------- help forpredxcat-------------------------------------------------------------------------------

Predicted means, medians, or proportions for nominal X's

predxcatyvar[ifexp] [inrange],xvar(xvar1[xvar2])[adjust(covlist)modellevel(#)linearmediangraphbarcluster(cluster_var)savepred(filename)graph_options]

Description

predxcatcalculates and optionally graphs means from linear regression models, medians from quantile regression models, or proportions from logistic regression models for one or two nominal X variables, adjusted for covariates. If a second X is specified, means, medians or proportions are calculated for all possible combinations of X categories, and an interaction effect is tested. Optionally, model estimates and/or a graph may be displayed. Dummy variables are created forxvar1. The lowest dummy variable defaults to the reference group. Ifxvar2is specified, dummy variables are created for it, as well as interaction terms. (Alternative to adjust command.)

Variables and options required

yvaris the dependent variableIf

yvaris continuous, defaults to linear regressionIf

yvaris binary (0,1), defaults to logistic regression

xvar(xvar1)is the nominal variable for categories of estimated means, medians, or proportions

xvar(xvar1 xvar2)gives categories of all combinations ofxvar1andxvar2; interaction betweenxvar1andxvar2is tested (Partial F for linear or quantile regression and likelihood ratio test for logistic regression)

Options

adjust(covlist)lists any covariates. If none are specified, unadjusted means, medians, or proportions are reported. Covariates are set to their mean, based on observations used in the analysis, or can be set to user specified values (e.g., age=50 gender=1). (Note: if anadjustvariable is nominal with more than 2 categories, it must be defined with dummy variables in theadjustlist, whereas dummy variables are created automatically in thexvaroption)

model-- for display purposes only, this option prints the regression table

level(#)-- specifies the confidence level, in percent, for calculation of confidence intervals (default=95%)

graph-- if one X (xvar1), graphs means, medians, or proportions and confidence intervals; if bothxvar1andxvar2are specified, points are graphed for each mean, median, or proportion, but confidence intervals are not graphed;xvar1is used for the x-axis with separate points for categories ofxvar2

bar-- can be used with thegraphoption to display a bar graph instead of points

linear-- requests linear regression whenyvaris binary (0,1); if not specified, logistic regression is assumed

median-- requests quantile regression whenyvaris continuous

cluster(cluster_var)-- Name of cluster variable; adjusts standard errors for intraclass correlation

savepred(filename)-- saves adjusted values and CI's to a Stata file

Examples

. predxcat chol, xvar(race) adjust(sbp age=50) modelUses linear regression to calculate mean cholesterol level by race category, adjusted for mean sbp and age=50; displays model

. predxcat chol, xvar(ses) adjust(sbp age) graphUses linear regression to calculate mean cholesterol by levels of socio-economic status, adjusted for sbp and age; displays graph

. predxcat htn, xvar(gender race) adjust(age smoke etoh)Uses logistic regression to calculate proportion of hypertensives for all combinations of gender (2 categories) and race (4 categories) for a total of 8 proportions, adjusted for age, smoking status, and alcohol consumption; tests for interaction between gender and race

. predxcat htn, xvar(race gender) adjust(age smoke etoh) graph barUses logistic regression to calculate the adjusted proportion of hypertensives for all combinations of gender and race; bar graph of proportions, race on the x-axis

Also seeManual:

[U] 23 Estimation and post-estimation commands,[R] adjust

Author

J.M.Garrett, Professor, School of Medicine, University of North Carolina,