feval

Predict responses of generalized linear regression model using one input for each predictor

Syntax

ypred = feval(mdl,Xnew1,Xnew2,...,Xnewn)

Description

ypred = feval(mdl,Xnew1,Xnew2,...,Xnewn) returns the predicted response of mdl to the new input predictors Xnew1,Xnew2,...,Xnewn.

Examples

Predict Response Values

Create a generalized linear regression model, and plot its responses to a range of input data.

Generate sample data using Poisson random numbers with two underlying predictors X(:,1) and X(:,2).

rng('default') % For reproducibility
rndvars = randn(100,2);
X = [2 + rndvars(:,1),rndvars(:,2)];
mu = exp(1 + X*[1;2]);
y = poissrnd(mu);

Create a generalized linear regression model of Poisson data.

mdl = fitglm(X,y,'y ~ x1 + x2','Distribution','poisson');

Generate a range of values for X(:,1) and X(:,2), and plot the predictions at the values.

[Xtest1,Xtest2] = meshgrid(min(X(:,1)):.5:max(X(:,1)),min(X(:,2)):.5:max(X(:,2)));
Z = feval(mdl,Xtest1,Xtest2);
surf(Xtest1,Xtest2,Z)

Input Arguments

mdl — Generalized linear regression model

GeneralizedLinearModel object | CompactGeneralizedLinearModel object

Generalized linear regression model, specified as a GeneralizedLinearModel object created using fitglm or stepwiseglm, or a CompactGeneralizedLinearModel object created using compact.

Xnew1,Xnew2,...,Xnewn — New predictor input values

vector | matrix | table | dataset array

New predictor values, specified as a vector, matrix, table, or dataset array.

• If you pass multiple inputs Xnew1,Xnew2,...,Xnewn and each includes observations for one predictor variable, then each input must be a vector. Each vector must have the same size. If you specify a predictor variable as a scalar, then feval expands the scalar argument into a constant vector of the same size as the other arguments.

• If you pass a single input Xnew1, then Xnew1 must be a table, dataset array, or matrix.

  • If Xnew1 is a table or dataset array, it must contain predictors that have the same predictor names as in the PredictorNames property of mdl.

  • If Xnew1 is a matrix, it must have the same number of variables (columns) in the same order as the predictor input used to create mdl. Note that Xnew1 must also contain any predictor variables that are not used as predictors in the fitted model. Also, all variables used in creating mdl must be numeric. To treat numerical predictors as categorical, identify the predictors using the 'CategoricalVars' name-value pair argument when you create mdl.

Data Types: single | double | table

Output Arguments

ypred — Predicted response values

numeric vector

Predicted response values at Xnew1,Xnew2,...,Xnewn, returned as a numeric vector.

For a binomial model, feval uses 1 as the BinomialSize parameter, so the values in ypred are predicted probabilities. To return the numbers of successes in the trials, use the predict function and specify the number of trials by using the 'BinomialSize' name-value pair argument.

For a model with an offset, feval uses 0 as the offset value. To specify the offset value used when you fit a model, use the predict function and the 'Offset' name-value pair argument.

Tips

• A regression object is, mathematically, a function that estimates the relationship between the response and predictors. The feval function enables an object to behave like a function in MATLAB^®. You can pass feval to another function that accepts a function input, such as fminsearch and integral.

• feval can be simpler to use with a model created from a table or dataset array. When you have new predictor data, you can pass it to feval without creating a table or matrix.

Alternative Functionality

• predict gives the same predictions as feval if you use the default values for the 'Offset' and 'BinomialSize' name-value pair arguments of predict. The prediction values can be different if you specify other values for these arguments. The predict function also returns confidence intervals on its predictions. Note that the predict function accepts a single input argument containing all predictor variables, rather than multiple input arguments with one input for each predictor variable.

• random predicts responses with added noise.