Scribd
Upload
18 views

Likelihood Ratio Tests

Uploaded by

datjacksonchan
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
18 views

Likelihood Ratio Tests

Uploaded by

datjacksonchan
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

Likelihood Ratio Tests

Likelihood ratio tests (LRTs) have been used to compare two nested models. The form of the
test is suggested by its name,

LRT = –2 log/ Ž ^ •
^
_= ())
,
_1 ())

the ratio of two likelihood functions; the simpler model (s) has fewer parameters than the general
(g) model. Asymptotically, the test statistic is distributed as a chi-squared random variable, with
degrees of freedom equal to the difference in the number of parameters between the two
models.

Likelihood ratio tests compare two models provided the simpler model is a special case of the
more complex model (i.e., “nested"). LRTs can be presented as a difference in the log-
likelihoods (recall that log(A/B) = logA – logB) and this is often handy as they can be expressed
in terms of deviance. Then,

LRT = –2 Šlog/ (_s ) – log/ (_g )‹

= –2 log/ (_s ) + 2 log/ (_g )

= deviances – devianceg .

Thus, the LRT can be computed as a difference in the deviance for the two models (ignoring the
term for the saturated model). This is convenient as the deviance is a value of interest in other
respects.

Say we flipped two coins, we could have

Ho : p1 œ p2 ´ p (the simpler model, 1 parameter),

vs. the alternative of different probabilities of heads, hence

Ha : p1 Á p2 . (the more general model, 2 parameters).


Compute MLEs under both models and compute the deviances (D),

Ds œ • 2log/ (_(^p)) , Ks œ 1, the number of parameters

Dg œ • 2log/ (_(^p1 , ^p2)) , Kg œ 2, the number of parameters.

The the LRT œ Ds • Dg , df œ Kg • K s œ 1.

Thus, this test statistic in approximately ;# with 1 df under the null hypothesis. The
approximation improves as sample size increases. Note, too that the log-likelihood for the
saturated model is a constant and the same for both of the above models; thus it was deleted in
this example.

Testing of null hypotheses has seen decreasing use in many areas of applied science over the
past 2 decades. We will made some reference to LRTs so that students can better understand
existing literature that makes use of these methods. Problems with statistical hypothesis testing
will be outlined at a latter point.

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy