How do I get a group mean holding for all other variables (using multiple regression)? I saw an analysis that did this, and am trying to produce something similar for a different set of data.
For example, using the Prestige dataset from the car pacakge:
library(car)
df<-Prestige
df$Group<-ifelse(df$women>.25,"High","Low") #this is a useless variable for regression, but I put this in because the real data i'm working with has multiple categorical variables, which makes it more confusing (I know how to get the means when there are only continuous vars)
reg<-lm(income~education+women+prestige+census+factor(type)+factor(Group),data=df)
summary(reg)
gives me the following output
Call:
lm(formula = income ~ education + women + prestige + census +
factor(type) + factor(Group), data = df)
Residuals:
Min 1Q Median 3Q Max
-7743.7 -947.9 -331.8 744.8 14307.5
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) -33.79732 3056.02530 -0.011 0.991201
education 130.75186 290.21943 0.451 0.653414
women -52.78426 10.00398 -5.276 9.03e-07 ***
prestige 139.42427 36.59482 3.810 0.000254 ***
census 0.04043 0.23694 0.171 0.864892
factor(type)prof 534.53024 1810.15685 0.295 0.768449
factor(type)wc 368.17807 1181.93287 0.312 0.756137
factor(Group)Low 367.09089 1274.25821 0.288 0.773946
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Residual standard error: 2646 on 90 degrees of freedom
(4 observations deleted due to missingness)
Multiple R-squared: 0.6366, Adjusted R-squared: 0.6084
F-statistic: 22.53 on 7 and 90 DF, p-value: < 2.2e-16Coefficients:
I know that 368.17807 is the difference between the means between type==wc and the reference group (type==bc), but how do I get the actual means? This should be DIFFERENT than just calculating the mean of all observations with each type, which can be found by
aggregate(income~type,data=df,FUN=mean)