predict.scoutobject {scout} | R Documentation |
A function to perform prediction, using an x matrix and the output of the "scout" function.
predict.scoutobject(object, newx, ...)
object |
The results of a call to the "scout" function. The coefficients that are part of this object will be used for making predictions. |
newx |
The new x at which predictions should be made. Can be a vector of length ncol(x), where x is the data on which scout.obj was created, or a matrix with ncol(x) columns. |
... |
Additional arguments to predict |
yhat |
If newx was a vector, then a matrix will be returned, with dimension length(lam1s)xlength(lam2s) (where lam1s and lam2s are attributes of scout.obj). The (i,j) element of this matrix will correspond to tuning parameter values (lam1s[i], lam2s[j]). If newx is a matrix, then an array of dimension nrow(newx)xlength(lam1s)xlength(lam2s) will be returned. |
Daniela M. Witten and Robert Tibshirani
Witten, DM and Tibshirani, R (2008) Covariance-regularized regression and classification for high-dimensional problems. {it Journal of the Royal Statistical Society, Series B}. To appear. <http://www-stat.stanford.edu/~dwitten>
data(diabetes) attach(diabetes) # Split data into training and test set training <- sample(nrow(x2),floor(nrow(x2)/2)) xtrain <- x2[training,] ytrain <- y[training] xtest <- x2[-training,] ytest <- y[-training] # Done splitting data into training and test set # Do cross-validation to determine best tuning parameter values for Scout(1,1) cv.out <- cv.scout(xtrain,ytrain,p1=1,p2=1, lam1s=seq(0.001,.15,len=10)) print(cv.out) # Done cross-validation # Fit Model scout.object <- scout(xtrain,ytrain,p1=1,p2=1,lam1s=cv.out$bestlam1,lam2s=cv.out$bestlam2) print(scout.object) # Done Fitting Model # Predict on test data, and report MSE yhats <- predict(scout.object,xtest) print(mean((yhats-ytest)^2)) detach(diabetes)