test_multivariate_zip_half_zeros() sometimes fails due to inaccuracy, depending on choice of RNG seed

[certifiedwaif]

set.seed(5)
options(recover = dump.frames)
test_univariate_zip()
# TODO: Add some sort of test for the accuracy of the approximation?

# Tests with multivariate fixed effects
test_multivariate_zip_no_zeros()
test_multivariate_zip_half_zeros()

produces the error

Error: result_var$a_rho/(result_var$a_rho + result_var$b_rho) not equal to expected_rho Mean relative difference: 0.2394163

[JohnOrmerod]

Are you around? Do you want to discuss?

---

From: Mark Greenaway [notifications@github.com] Sent: Friday, June 06, 2014 3:43 PM To: certifiedwaif/phd Subject: [phd] test_multivariate_zip_half_zeros() sometimes fails due to inaccuracy, depending on choice of RNG seed (#5)

set.seed(5) options(recover = dump.frames) test_univariate_zip() # TODO: Add some sort of test for the accuracy of the approximation? # Tests with multivariate fixed effects test_multivariate_zip_no_zeros() test_multivariate_zip_half_zeros()

produces the error

Error: result_var$a_rho/(result_var$a_rho + result_var$b_rho) not equal to expected_rho Mean relative difference: 0.2394163

— Reply to this email directly or view it on GitHubhttps://github.com/certifiedwaif/phd/issues/5.

[certifiedwaif]

Hello John,

Yes, I'm around. At the moment, I'm working on the RWMH for the model, having just finished the probit regression question from last year's assignment.

I guess we could look at why the test fails in that particular case. I just wanted to note the issue down so that I don't forget about it.

Mark

On Fri, Jun 6, 2014 at 3:46 PM, JohnOrmerod notifications@github.com wrote:

Are you around? Do you want to discuss?

---

From: Mark Greenaway [notifications@github.com] Sent: Friday, June 06, 2014 3:43 PM To: certifiedwaif/phd Subject: [phd] test_multivariate_zip_half_zeros() sometimes fails due to inaccuracy, depending on choice of RNG seed (#5)

set.seed(5) options(recover = dump.frames) test_univariate_zip() # TODO: Add some sort of test for the accuracy of the approximation? # Tests with multivariate fixed effects test_multivariate_zip_no_zeros() test_multivariate_zip_half_zeros()

produces the error

Error: result_var$a_rho/(result_var$a_rho + result_var$b_rho) not equal to expected_rho Mean relative difference: 0.2394163

— Reply to this email directly or view it on GitHub< https://github.com/certifiedwaif/phd/issues/5>.

— Reply to this email directly or view it on GitHub https://github.com/certifiedwaif/phd/issues/5#issuecomment-45304881.