-
Notifications
You must be signed in to change notification settings - Fork 2k
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
GH-15891: correct sign in AIC calculation as the negative loglikeliho…
…od is used (#15932) * GH-15891: correct sign in AIC calculation as the negative loglikelihood is used * GH-15891: add test for AIC * GH-15891: correct test for AIC * GH-15891: fix old tests --------- Co-authored-by: syzonyuliia <[email protected]>
- Loading branch information
1 parent
965a03c
commit 79d8f67
Showing
4 changed files
with
96 additions
and
11 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
85 changes: 85 additions & 0 deletions
85
h2o-r/tests/testdir_algos/glm/runit_gh_15891_tweedie_aic.R
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,85 @@ | ||
| setwd(normalizePath(dirname(R.utils::commandArgs(asValues=TRUE)$"f"))) | ||
| source("../../../scripts/h2o-r-test-setup.R") | ||
|
|
||
| test_glm_tweedies <- function() { | ||
| require(statmod) | ||
| require(tweedie) | ||
| if (requireNamespace("tweedie")) { | ||
| num_rows <- 1000 | ||
| num_cols <- 5 | ||
| f1 <- random_dataset_real_only(num_rows, num_cols) # generate dataset containing the predictors. | ||
| f1R <- as.data.frame(h2o.abs(f1)) | ||
| weights <- c(0.1, 0.2, 0.3, 0.4, 0.5, 1) # weights to generate the mean | ||
| mu <- generate_mean(f1R, num_rows, num_cols, weights) | ||
| pow <- c(1.1) # variance power range | ||
| phi <- c(1) # dispersion factor range | ||
| y <- "resp" # response column | ||
| x <- c("abs.C1.", "abs.C2.", "abs.C3.", "abs.C4.", "abs.C5.") | ||
| for (ind in c(1:length(pow))) { # generate dataset with each variance power and dispersion factor | ||
| trainF <- generate_dataset(f1R, num_rows, num_cols, pow[ind], phi[ind], mu) | ||
| print(paste("Compare H2O, R GLM model coefficients and standard error for var_power=", pow[ind], "link_power=", 1 - pow[ind], sep = " ")) | ||
| compareH2ORGLM(pow[ind], 1 - pow[ind], x, y, trainF, as.data.frame(trainF), phi[ind]) | ||
| } | ||
| } else { | ||
| print("test_glm_tweedies is skipped. Need to install tweedie package.") | ||
| } | ||
| } | ||
|
|
||
| generate_dataset <- function(f1R, numRows, numCols, pow, phi, mu) { | ||
| resp <- tweedie::rtweedie(numRows, xi = pow, mu, phi, power = pow) | ||
| f1h2o <- as.h2o(f1R) | ||
| resph2o <- as.h2o(as.data.frame(resp)) | ||
| finalFrame <- h2o.cbind(f1h2o, resph2o) | ||
| return(finalFrame) | ||
| } | ||
|
|
||
| generate_mean <- function(f1R, numRows, numCols, weights) { | ||
| y <- c(1:numRows) | ||
| for (rowIndex in c(1:numRows)) { | ||
| tempResp = 0.0 | ||
| for (colIndex in c(1:numCols)) { | ||
| tempResp = tempResp + weights[colIndex] * f1R[rowIndex, colIndex] | ||
| } | ||
| y[rowIndex] = tempResp | ||
| } | ||
| return(y) | ||
| } | ||
|
|
||
| compareH2ORGLM <- | ||
| function(vpower, lpower, x, y, hdf, df, truedisp, tolerance = 2e-4) { | ||
| print("Define formula for R") | ||
| formula <- (df[, "resp"] ~ .) | ||
| rmodel <- glm( | ||
| formula = formula, | ||
| data = df[, x], | ||
| family = tweedie(var.power = vpower, link.power = | ||
| lpower), | ||
| na.action = na.omit | ||
| ) | ||
| rAIC <- AICtweedie(rmodel) | ||
| h2omodel <- | ||
| h2o.glm( | ||
| x = x, | ||
| y = y, | ||
| training_frame = hdf, | ||
| family = "tweedie", | ||
| link = "tweedie", | ||
| tweedie_variance_power = vpower, | ||
| tweedie_link_power = lpower, | ||
| alpha = 0.5, | ||
| lambda = 0, | ||
| nfolds = 0, | ||
| compute_p_values = TRUE, | ||
| calc_like = TRUE, | ||
| fix_tweedie_variance_power = TRUE | ||
| ) | ||
| h2oAIC <- h2o.aic(h2omodel) | ||
| print("Comparing H2O and R GLM model AIC.") | ||
| print("R AIC") | ||
| print(rAIC) | ||
| print("h2o model AIC") | ||
| print(h2oAIC) | ||
| expect_true(abs(rAIC - h2oAIC) / h2oAIC < 1e-2) | ||
| } | ||
|
|
||
| doTest("Comparison of H2O to R TWEEDIE family AIC with tweedie dataset", test_glm_tweedies) |