wilksLambdaTest: wilksLambdaTest

In mrparker909/mvregress: Classic Multivariate Linear Regression

Description

Performs Wilks likelihood ratio test on the multivariate regression model lm_mod versus the null hypothesis model lm_nullmod. Note that if n_adj=0 (the default) then the return values pval and pvalAdj will be the same. Otherwise, pval will be the unadjusted pvalue (full degrees of freedom), and pvalAdj will be the adjusted pvalue (reduced degrees of freedom).

Usage

wilksLambdaTest(lm_mod, lm_nullmod = NULL, n_adj = 0)

Arguments

lm_mod	fitted model (eg: lm(Y~X)) corresponding to the alternative hypothesis.
lm_nullmod	fitted model (eg: lm(Y~1)) corresponding to the null hypothesis, default is lm(Y~1) when lm_nullmod==NULL.
n_adj	default=0, subtracted from the error degrees of freedom (for example, may be used to correct for previous testing done on the data).

Examples

lam = NULL
pval = NULL
for(i in 1:1000) {
x1 = rnorm(100)
x2 = rnorm(100)
y1 = rnorm(100)
y2 = rnorm(100)
A = cbind(y1,y2)
B = cbind(x1,x2)
wi = wilksLambdaTest(lm(A~B), n_adj=0)
lam = c(lam, wi$lambda)
pval = c(pval, wi$pval)
}

hist(lam, freq=F, breaks=100)
curve(dchisq(x, wi$chisq_df), 0, max(lam), lwd=2, xlab = "", ylab = "", add = T)
hist(pval, freq=F, breaks=100)

lam = NULL
pval = NULL
for(i in 1:1000) {
x1 = rnorm(100)
x2 = rnorm(100)
y1 = rnorm(100)
y2 = rnorm(100)
y3 = rnorm(100)
y4 = rnorm(100)
y5 = rnorm(100)
y6 = rnorm(100)
A = cbind(y1,y2,y3,y4,y5,y6)
B = cbind(x1,x2)
wi = wilksLambdaTest(lm(A~B), n_adj=0)
lam = c(lam, wi$lambda)
pval = c(pval, wi$pval)
}

hist(lam, freq=F, breaks=100)
curve(dchisq(x, wi$chisq_df), 0, max(lam), lwd=2, xlab = "", ylab = "", add = T)
hist(pval, freq=F, breaks=100)

mrparker909/mvregress documentation built on Dec. 27, 2019, 3:40 p.m.