likelihood ratio

(redirected from Likelihood-ratio test)
Also found in: Encyclopedia, Wikipedia.

likelihood ratio

usually preceded by "maximum" (that is, maximum likelihood ratio), this ratio maximizes the probability that the parameters in the ratio agree with the empirically observed data.

like·li·hood ra·ti·o

(līk'lē-hud rā'shē-ō)
The ratio of the probability of a test result among patients with a certain disease or disorder to the probability of that same test result among patients who do not have the targeted disease or disorder.

likelihood ratio

(līk′lē-hood″),

LR

A statistical tool used to help determine the usefulness of a diagnostic test for including or excluding a particular disease. An LR = 1 suggests that the test ordered neither helps to diagnose the disease in question nor helps to rule it out. Higher LRs increase the probability that the disease will be present; LRs < 1.0 decrease the probability that the disease is present.

A positive LR can be thought of as the probability that someone with a suspected condition will, accurately, have a positive test result, divided by the probability that a healthy person will, inaccurately, test positive for the disease. Mathematically this can be represented by the following equation: LR+ = sensitivity of the test/ (1− specificity of the test). A negative LR is the probability that a sick person will fail to be detected by the test, divided by the probability that a healthy person will be accurately shown by the test to have no sign of disease. Mathematically: LR− = (1 − sensitivity of the test) / specificity of the test.

likelihood ratio

The percentage of ill people with a given test result divided by the percentage of well people with the same result. Ratios near unity should not influence decisions. This useful guide to refining clinical diagnosis is little used mainly because of its complexity; The Fagan nomogram can simplify the matter.
References in periodicals archive ?
Test m = 1 Statistic p-value Lagrange multiplier (LM) 5.33 0.000 (***) Likelihood-ratio test (LR) 21.54 0.000 (***) Test m = 2 Statistic p-value Lagrange multiplier (LM) 2.67 0.020 (**) Likelihood-ratio test (LR) 18.50 0.010 (**) Note: (***), (**) denote significance at 1% and 5%, respectively.
The likelihood-ratio test in all cases rejects the null hypothesis of no overdispersion (p value = .000).
(6) Hausman test (a) 0.04 [0.8462] Pcsaran's test (b) 6.510 [0.000] Frees' test (b) 1.455 {0.1078, 0.1408, 0.2034} (e) Friedman's test (b) 74.682 [0.0001 Likelihood-ratio test (d) 603.24 [0.000] Wooldridge test (e) 130.456 [0.000] Tests Hall-Roeger approach: Hall-Roeger cross- sectional model, Eq.
Likelihood-ratio tests were therefore carried out to determine whether the model allowing [beta] to vary freely yielded significantly better fits than one with three parameters and [beta] fixed at 3.2.
In practice this involves estimating all three of the models outlined above and conducting Johansen's likelihood-ratio tests for the rank order of the long-run matrix sequentially from the most restrictive to the least restrictive specification.
To investigate the properties of the likelihood-ratio test of convexity, we will compare it to the results of a log-log unit slope test applied to synthetic data (with various degrees of contamination of measurement error).
As each new set of parameters was added, a likelihood-ratio test was performed to determine whether the simpler model could be rejected (Goldman 1993; Yang 1996b).
Obviously, the likelihood-ratio test does not call for constrained ML estimation.
Meade and Lautenschlager (2004) provided a detailed comparison of the likelihood-ratio test within IRT and MACS1 that included a simulation study that examined the Type I error rate and power of both methods.
Likelihood-Ratio Test of Identical Speciation Times
In this case, a measure is needed of the average selection pattern of the group, where data from each individual are weighted equally instead of weighting according to the number of observations as is done by the likelihood-ratio test. This can be accomplished by treating the values of [Mathematical Expression Omitted] as random variables, with the set of index values for each animal representing an independent observation.
A likelihood-ratio test for this model against the null hypothesis of no selection on either trait (i.e., [Beta] = 0 for each trait) indicated a significant effect of the traits on survival ([Mathematical Expression Omitted], P = 0.017).