Predict using a cv.ptLasso object.

Return predictions and performance measures for a test set.

Usage

# S3 method for class 'cv.ptLasso'
predict(
  object,
  xtest,
  groupstest = NULL,
  ytest = NULL,
  alpha = NULL,
  alphatype = c("fixed", "varying"),
  type = c("link", "response", "class"),
  s = "lambda.min",
  gamma = "gamma.min",
  return.link = FALSE,
  ...
)

Arguments

object: Fitted "cv.ptLasso" object.
xtest: Input matrix, matching the form used by "cv.ptLasso" for model training.
groupstest: A vector indicating to which group each observation belongs. Coding should match that used for model training. Will be NULL for target grouped data.
ytest: Response variable. Optional. If included, "predict" will compute performance measures for xtest using "type.measure" from the cvfit object.
alpha: The chosen alpha to use for prediction. May be a vector containing one value of alpha for each group. If NULL, this will rely on the choice of "alphatype".
alphatype: Choice of '"fixed"' or '"varying"'. If '"fixed"', use the alpha that achieved best cross-validated performance. If '"varying"', each group uses the alpha that optimized the group-specific cross-validated performance.
type: Type of prediction required. Type '"link"' gives the linear predictors for '"binomial", '"multinomial"' or '"cox"' models; for '"gaussian"' models it gives the fitted values. Type '"response"' gives the fitted probabilities for '"binomial"' or '"multinomial"', and the fitted relative-risk for '"cox"'; for '"gaussian"' type '"response"' is equivalent to type '"link"'. Note that for '"binomial"' models, results are returned only for the class corresponding to the second level of the factor response. Type '"class"' applies only to '"binomial"' or '"multinomial"' models, and produces the class label corresponding to the maximum probability.
s: Value of the penalty parameter 'lambda' at which predictions are required. Will use the same lambda for all models; can be a numeric value, '"lambda.min"' or '"lambda.1se"'. Default is '"lambda.min"'.
gamma: For use only when 'relax = TRUE' was specified during training. Value of the penalty parameter 'gamma' at which predictions are required. Will use the same gamma for all models; can be a numeric value, '"gamma.min"' or '"gamma.1se"'. Default is '"gamma.min"'.
return.link: If TRUE, will additionally return the linear link for the overall, pretrained and individual models: linkoverall, linkpre and linkind.
...: other arguments to be passed to the "predict" function.

Value

A list containing the requested predictions. If ytest is included, will also return error measures.

call: The call that produced this object.
alpha: The value(s) of alpha used to generate predictions.
yhatoverall: Predictions from the overall model.
yhatind: Predictions from the individual models.
yhatpre: Predictions from the pretrained models.
supoverall: Indices of the features selected by the overall model.
supind: Union of the indices of the features selected by the individual models.
suppre.common: Features selected in the first stage of pretraining.
suppre.individual: Union of the indices of the features selected by the pretrained models, without the features selected in the first stage.
type.measure: If ytest is supplied, the performance measure computed.
erroverall: If ytest is supplied, performance for the overall model. This is a named vector containing performance for (1) the entire dataset, (2) the average performance across groups, (3) the average performance across groups weighted by group size and (4) group-specific performance.
errind: If ytest is supplied, performance for the overall model. As described in erroverall.
errpre: If ytest is supplied, performance for the overall model. As described in erroverall.
linkoverall: If return.link is TRUE, return the linear link from the overall model.
linkind: If return.link is TRUE, return the linear link from the individual models.
linkpre: If return.link is TRUE, return the linear link from the pretrained models.

Author

Erin Craig and Rob Tibshirani
Maintainer: Erin Craig <erincr@stanford.edu>

Examples

#### Gaussian example
set.seed(1234)
out = gaussian.example.data(k=2, class.sizes = c(50, 50))
x = out$x; y=out$y; groups = out$group;
outtest = gaussian.example.data(k=2, class.sizes = c(50, 50))
xtest=outtest$x; ytest=outtest$y; groupstest=outtest$groups;

# Model fitting
# By default, use the single value of alpha that had the best CV performance on the entire dataset:
cvfit = cv.ptLasso(x, y, groups = groups, family = "gaussian", type.measure = "mse")
pred = predict(cvfit, xtest, groupstest, ytest=ytest, s="lambda.min")
pred
#> 
#> Call:  
#> predict.cv.ptLasso(object = cvfit, xtest = xtest, groupstest = groupstest,  
#>     ytest = ytest, s = "lambda.min") 
#> 
#> 
#> alpha =  0.5 
#> 
#> Performance (Mean squared error):
#> 
#>            allGroups  mean group_1 group_2    r^2
#> Overall        609.0 609.0   548.3   669.7 0.2330
#> Pretrain       650.8 650.8   601.1   700.6 0.1803
#> Individual     675.2 675.2   667.6   682.8 0.1496
#> 
#> Support size:
#>                                         
#> Overall    21                           
#> Pretrain   18 (6 common + 12 individual)
#> Individual 26                           

# For each group, use the value of alpha that had the best CV performance for that group:
pred = predict(cvfit, xtest, groupstest, ytest=ytest, s="lambda.min", alphatype = "varying")
pred
#> 
#> Call:  
#> predict.cv.ptLasso(object = cvfit, xtest = xtest, groupstest = groupstest,  
#>     ytest = ytest, alphatype = "varying", s = "lambda.min") 
#> 
#> 
#> 
#> alpha:
#> [1] 0.1 1.0
#> 
#> 
#> Performance (Mean squared error):
#>            overall  mean wtdMean group_1 group_2
#> Overall      609.0 609.0   609.0   548.3   669.7
#> Pretrain     629.7 629.7   629.7   576.6   682.8
#> Individual   675.2 675.2   675.2   667.6   682.8
#> 
#> 
#> Support size:
#>                                         
#> Overall    21                           
#> Pretrain   26 (6 common + 20 individual)
#> Individual 26                           

# Specify a single value of alpha and use lambda.1se.
pred = predict(cvfit, xtest, groupstest, ytest=ytest, s="lambda.1se",
               alphatype = "varying", alpha = .3)
pred
#> 
#> Call:  
#> predict.cv.ptLasso(object = cvfit, xtest = xtest, groupstest = groupstest,  
#>     ytest = ytest, alpha = 0.3, alphatype = "varying", s = "lambda.1se") 
#> 
#> 
#> 
#> alpha =  0.3 
#> 
#> Performance (Mean squared error):
#> 
#>            allGroups  mean group_1 group_2     r^2
#> Overall        709.1 709.1   573.8   844.5 0.10690
#> Pretrain       660.6 660.6   585.7   735.5 0.16800
#> Individual     743.8 743.8   667.6   820.0 0.06322
#> 
#> Support size:
#>                                       
#> Overall    6                          
#> Pretrain   7 (6 common + 1 individual)
#> Individual 10                         

# Specify a vector of choices for alpha: 
pred = predict(cvfit, xtest, groupstest, ytest=ytest, s="lambda.min",
               alphatype = "varying", alpha = c(.1, .5))
pred
#> 
#> Call:  
#> predict.cv.ptLasso(object = cvfit, xtest = xtest, groupstest = groupstest,  
#>     ytest = ytest, alpha = c(0.1, 0.5), alphatype = "varying",  
#>     s = "lambda.min") 
#> 
#> 
#> alpha:
#> [1] 0.1 0.5
#> 
#> 
#> Performance (Mean squared error):
#>            overall  mean wtdMean group_1 group_2
#> Overall      609.0 609.0   609.0   548.3   669.7
#> Pretrain     638.6 638.6   638.6   576.6   700.6
#> Individual   675.2 675.2   675.2   667.6   682.8
#> 
#> 
#> Support size:
#>                                         
#> Overall    21                           
#> Pretrain   18 (6 common + 12 individual)
#> Individual 26

Usage

Arguments

Value

See also

Author

Examples