CosmoPhotoz (version 0.1)

glmTrainPhotoZ: Fit a glm for photometric redshift estimation

Description

glmTrainPhotoZ trains a generalized linear model for photometric redshift estimation.

Usage

glmTrainPhotoZ(x, formula, method, family)

Arguments

x

a data.frame containing the data to train the model

formula

an object of class "formula" to be adopted

method

a string containing the chosen glm method. Two options are available: Frequentist will use the function glm from the package stats; Bayesian will use the function bayesglm from the package arm.

family

a string containing gamma or inverse.gaussian (a string with a description of the error distribution and link function to be used in the model).

Value

a trained glm object containing the fit of the model

Details

The program is a simple alteration of glm() and bayesglm(). The new arguments here are: Frequentist, Bayesian.

Examples

## Not run: 
# # Load the data
# data(PHAT0train)
# data(PHAT0test)
# 
# # Combine the training and test data and calculate the principal components
# PC_comb <- computeCombPCA(subset(PHAT0train, select=c(-redshift)),
#            subset(PHAT0test, select=c(-redshift)),
#            robust=FALSE) # robust is false here just to make it faster
# Trainpc <- cbind(PC_comb$x, redshift=PHAT0train$redshift)
# Testpc <- PC_comb$y
# 
# # Fitting
# Fit <- glmTrainPhotoZ(Trainpc, formula=redshift~poly(Comp.1,2)*
#            poly(Comp.2,2)*Comp.3*Comp.4*Comp.5*Comp.6,
#            method="Bayesian", family="gamma")
# 
# # Perform the photometric redshift estimation
# photoz <- predict(Fit$glmfit, newdata=Testpc, type="response")
# specz <- PHAT0test$redshift
# 
# # Show a plot with the results
# plotDiagPhotoZ(photoz, specz, "box")
# ## End(Not run)