Add many single terms to a model {Rfast2}

Add many single terms to a model

Description

Add many single terms to a model.

Usage

add.term(y, xinc, xout, devi_0, type = "logistic", logged = FALSE,
tol = 1e-07, maxiters = 100, parallel = FALSE)    

Arguments

y

The response variable. It must be a numerical vector.

xinc

The already included indendent variable(s).

xout

The independent variables whose conditional association with the response is to be calculated.

devi_0

The deviance for Poisson, logistic, qpoisson, qlogistic and normlog regression or the log-likelihood for the Weibull, spml and multinomial regressions. See the example to understand better.

type

The type of regression, "poisson", "logistic", "qpoisson" (quasi Poisson), "qlogistic" (quasi logistic) "normlog" (Gaussian regression with log-link) "weibull", "spml" and "multinom".

logged

Should the logarithm of the p-value be returned? TRUE or FALSE.

tol

The tolerance value to terminate the Newton-Raphson algorithm when fitting the regression models.

maxiters

The maximum number of iterations the Newton-Raphson algorithm will perform.

parallel

Should the computations take place in parallel? TRUE or FALSE.

Details

The function is similar to the built-in function add1. You have already fitted a regression model with some independent variables (xinc). You then add each of the xout variables and test their significance.

Value

A matrix with two columns. The test statistic and its associated (logged) p-value.

Author(s)

Stefanos Fafalios.

R implementation and documentation: Stefanos Fafalios <[e-mail]>.

References

McCullagh, Peter, and John A. Nelder. Generalized linear models. CRC press, USA, 2nd edition, 1989.

Presnell Brett, Morrison Scott P. and Littell Ramon C. (1998). Projected multivariate linear models for directional data. Journal of the American Statistical Association, 93(443): 1068-1077.

See Also

bic.regs, logiquant.regs, sp.logiregs

Examples

x <- matrix( rnorm(200 * 10), ncol = 10)
y <- rpois(200, 10)
devi_0 <- deviance( glm(y ~ x[, 1:2], poisson) )
a <- add.term(y, xinc = x[,1:2], xout = x[, 3:10], devi_0 = devi_0, type= "poisson")

y <- rbinom(200, 1, 0.5)
devi_0 <- deviance( glm(y ~ x[, 1:2], binomial) )
a <- add.term(y, xinc = x[,1:2], xout = x[, 3:10], devi_0 = devi_0, type= "logistic")


y <- rbinom(200, 2, 0.5)
devi_0 <- Rfast::multinom.reg(y, x[, 1:2])$loglik
a <- add.term(y, xinc = x[,1:2], xout = x[, 3:10], devi_0 = devi_0, type= "multinom")

y <- rgamma(200, 3, 1)
devi_0 <- Rfast::weib.reg(y, x[, 1:2])$loglik
a <- add.term(y, xinc = x[,1:2], xout = x[, 3:10], devi_0 = devi_0, type= "weibull")