lmc: Linear Model Fitting with Constraints

In GJRM: Generalised Joint Regression Modelling

Linear Model Fitting with Constraints

Description

Linear model fitting with positivity and sum-to-one constraints on the model's coefficients.

Usage

lmc(y, X, start.v = NULL, lambda = 1, pen = "none", gamma = 1, a = 3.7) 
        
        

Arguments

y	Response vector.
X	Design matrix.
start.v	Starting values.
lambda	Tuning parameter.
pen	Type of penalty. Choices are: none, ridge, lasso, alasso, scad.
gamma	Power parameter of adaptive lasso.
a	Scad parameter.

Details

Linear model fitting with positivity and sum-to-one constraints on the model's coefficients.

Value

The function returns an object of class lmc.

Examples

## Not run:  

library(GJRM)

set.seed(1)

n    <- 1000
beta <- c(0.07, 0.08, 0.21, 0.12, 0.15, 0.17, 0.2)
l    <- length(beta)
X    <- matrix(runif(n*l), n, l)

y    <- X%*%beta + rnorm(n)

out <- lmc(y, X)
conv.check(out)

out1 <- lmc(y, X, start.v = beta)
conv.check(out1)


coef(out)                    # estimated   coefficients
round(out$c.coefficients, 3) # constrained coefficients
sum(out$c.coefficients)

round(out1$c.coefficients, 3) 
sum(out1$c.coefficients)


# penalised estimation

out1 <- lmc(y, X, pen = "alasso", lambda = 0.02)
conv.check(out1)

coef(out1)                    
round(out1$c.coefficients, 3)
sum(out1$c.coefficients)


AIC(out, out1)
BIC(out, out1)

round(cbind(out$c.coefficients, out1$c.coefficients), 3)

# scad

n    <- 10000
beta <- c(0.2, 0, 0, 0.02, 0.01, 0.01, 0.01, 0.08, 0.21, 0.12, 0.15, 0.17, 0.02)
l    <- length(beta)
X    <- matrix(runif(n*l), n, l)

y    <- X%*%beta + rnorm(n)

out1 <- lmc(y, X, pen = "scad", lambda = 0.01)
conv.check(out1)

coef(out1)  
sum(out1$c.coefficients)
                  
round(cbind(beta, out1$c.coefficients), 2)


## End(Not run)

GJRM documentation built on July 9, 2023, 7:15 p.m.