0

I am trying to compute the log likelihood ratio test in R, but am having some difficulties.

For some reason I keep getting a negative log likelihood value which isn't possible I do not know the reason.

This is the data I am using.

Here is the code so far:

I am trying to test the null hypothesis that the mean is not equal to (1,1,1) The reason I multiply the covariance matrix by (n4-1)/n4 is that I need to divide the covariance by n, not n-1, and the cov function divides the matrix by n-1.

data <- read.csv('dat1.csv')
data <- data[, 2:4]
datamat <- as.matrix(data, nrow=25, ncol=3)

mu0_4 <- c(1,1,1)
n4 <- dim(datamat)[1]

xbar4 <- colMeans(datamat)
hs4 <- cov(datamat - xbar4)*(n4-1)/n4
det_hs4 <- det(hs4)
det_hs4

hs04 <- cov(datamat - mu0_4)*(n4-1)/n4
det_hs04 <- det(hs04)
det_hs04

LRS <- (det_hs4/det_hs04)^(n4/2)
l_lrs <- -2*log(LRS)
l_lrs

I am unsure of the reason I am getting a negative value, but if someone could please give me some advice that would be much appreciated.

Thank you for reading

mrsquid
  • 605
  • 2
  • 9
  • 24

1 Answers1

0

I think you are making a mistake in the code. In (datamat - xbar4) the subtraction happens in row-wise. Try 

sweep(datamat, 2, xbar4 , "-")

I do not know why you said negative log-likelihood values cannot be negative. Usually, they are negative. Maybe you meant the LRS.

kangaroo_cliff
  • 6,067
  • 3
  • 29
  • 42