The following model looks at factors that influence the use of prescription medications

for anxiety and depression using data from the public use file of the $2012$ Canadian

Community Health Survey. The variable MEFDRUG is a binary variable which takes a

value of $1$ if the individual took a medication in the past $2$ days and $0$ otherwise. Table

Table $1$ describe the variables.

Table $1$: Variable Descriptions

The following $R$ code estimates $P($MEDFDRUG $=$ Yes $|$ MED$\_05,$ DHH$\_$SEX,

DHHAGE, GEN$\_01,$ SPSDCON, INCGHH,EDUR$04)$ using the LPM and the pro$-$

bit model and plots the two against DHHAGE $($age group$).$ Comment on the results

and discuss whether you think it is appropriate to use the LPM in this case.

$rm($ list $=1s())$

# Read in the CCHS data

mydata read.$csv(cchs-82M0013-E-2012-$mental$-healt{h}_{F}1.csv,$

header$=$TRUE$)$

# Create a subset with the necessary variables and remove any missing observations

train.data subset $($mydata$,$

mydata$MED &

mydata$MEDFDRUG $2$ &

mydata $ DHH$\_$SEX $2$ &

mydata$DHHGAGE &

mydata$GEN &

mydata$SPSDCON $40$ &

mydata $ INCGHH $5$ &

mydata$EDUDR

attach$($train$.$data$)$

# Redefine MEDFDRUG $=\{0,1\}$ instead of $\{1,2\}$

MEDFDRUG $[$MEDFDRUG $==2$

# Create a working dataset

mydata data.frame $($MEDFDRUG$,$

MED$\_05=$ factor$($MED$\_05),$

DHH$\_$SEX $=$ factor DHH$\_$SEX ${?}_{-}$