Randomttest: Difference between revisions
Jump to navigation
Jump to search
imported>Scott (Created page with " ===Purpose=== Randomization t-test for evaluating residuals from two models. ===Synopsis=== :prob = randomttest(err_1,err_2,iter) ===Description=== Pairwise comparison betwe...") |
imported>Scott |
||
Line 11: | Line 11: | ||
Pairwise comparison between two sets of model residuals using a randomization of the sign of the differences. Output is the probability that the two sets of residuals are not different. | Pairwise comparison between two sets of model residuals using a randomization of the sign of the differences. Output is the probability that the two sets of residuals are not different. | ||
Based on the publication: | Based on the publication: ''Hilko van der Voet "Comparing the predictive accuracy of models using a simple randomization test", Chemometrics and Intelligent Laboratory Systems 25 (1994) 313-323.'' | ||
====Inputs==== | ====Inputs==== | ||
Line 21: | Line 20: | ||
====Outputs==== | ====Outputs==== | ||
* '''prob''' = Probability that residuals (and thus models) are not significantly different. | * '''prob''' = Probability that residuals (and thus models) are not significantly different. | ||
===See Also=== | ===See Also=== | ||
[[crossval]], [[signtest]], [[wilcoxon]] | [[crossval]], [[signtest]], [[wilcoxon]] |
Latest revision as of 15:57, 27 September 2011
Purpose
Randomization t-test for evaluating residuals from two models.
Synopsis
- prob = randomttest(err_1,err_2,iter)
Description
Pairwise comparison between two sets of model residuals using a randomization of the sign of the differences. Output is the probability that the two sets of residuals are not different.
Based on the publication: Hilko van der Voet "Comparing the predictive accuracy of models using a simple randomization test", Chemometrics and Intelligent Laboratory Systems 25 (1994) 313-323.
Inputs
- err_1 = Prediction errors from model #1
- err_2 = Prediction errors from model #2
Optional Inputs
- iter = Number of iterations to perform. Default = 199.
Outputs
- prob = Probability that residuals (and thus models) are not significantly different.