# retain top 20 variables
to.retain=20
imp.vars.energy=as.character(read.csv("gbm.importance.50.energy.csv",header=TRUE)[1:to.retain,"var"])
imp.vars.inj.type=as.character(read.csv("gbm.importance.50.code.csv",header=TRUE)[1:to.retain,"var"])
imp.vars.body=as.character(read.csv("gbm.importance.50.body.csv",header=TRUE)[1:to.retain,"var"])

imp.vars.severity=c("crane","heavy.material.tool","stairs","unpowered.tool","exiting.transitioning","heavy.vehicle","uneven.surface","drill","ladder","object.at.height","improper.procedure.inattention","unpowered.transporter","machinery","improper.security.of.materials","working.at.height","no.improper.PPE","small.particle","steel.steel.sections","stripping","scaffold")

# find column numbers corresponding to the important variables (where binary is some attribute and outcome data set)

## Read Body part data and select the important variables
data = read.csv("data.body.part.csv")
index=which(colnames(data)%in%imp.vars.body)

Xpredictors = data[,index]

bpart=data[,1]
bpart=as.character(bpart)
bpart[bpart == "head"]=1
bpart[bpart == "neck"]=2
bpart[bpart == "trunk"]=3
bpart[bpart == "upper extremities"]=4
bpart[bpart == "lower extremities"]=5

### create binary vector
bpartbin=as.numeric(bpart)

### Now fit Multinomial logistic regression using Xpredictors and ## bpartbin