nflcomb <- read.csv("http://www.stat.ufl.edu/~winner/sta4210/mydata/nfl_combine.csv",
       header=TRUE)
attach(nflcomb); names(nflcomb)

#### Obtain the Training and Validation Samples
#### nfl.samp obtains a sample of 168 integers from 1:335 w/out replacement
#### Selects all rows where id is either in nfl.samp (train) or not (valid) 
#### Selects all columns:     nflcomb[rows,cols]

set.seed(9876)   #### Will produce same sample in future runs
nfl.samp <- sample(1:length(Weight),168,replace=FALSE)
nfl.train <- nflcomb[nfl.samp,]
nfl.valid <- nflcomb[-nfl.samp,]

#### Fit the Regression model for the Training and Validation Samples

train.mod <- lm(Weight ~ Height + ArmLng + HandLng, data=nfl.train)
valid.mod <- lm(Weight ~ Height + ArmLng + HandLng, data=nfl.valid)
summary(train.mod)
summary(valid.mod)

#### Compute the fitted values for Validation sample from Model fitted
####   to Training Sample (Note that X1, X2, X3 are in Columns 4, 5, 6)
####   and Y is in Column 7
####  Compute MSEP

yhat.val <- predict(train.mod,nfl.valid[,4:6])
e.val <- nfl.valid[,7] - yhat.val
(MSEP <- sum(e.val^2)/nrow(nfl.valid))