search

jasp-stats / jasp-issues

This repository is solely meant for reporting of bugs, feature requests and other issues in JASP.
57 stars 29 forks source link

bayes failure for linear mixed #957

Closed kornbrot closed 3 years ago

kornbrot commented 3 years ago

Error in dimnamesGets(x, value): invalid dimnames given for “dgCMatrix” object

Stack trace

To receive assistance with this problem, please report the message above at: https://jasp-stats.org/bug-reports

Steps to reproduce:

  1. Go to analysis tab
  2. Click on '...bayes linear mixed.' dependent proptn fixed wiithin random pno'
  3. See error This analysis terminated unexpectedly. Error in dimnamesGets(x, value): invalid dimnames given for “dgCMatrix” object Stack trace

To receive assistance with this problem, please report the message above at: https://jasp-stats.org/bug-reports

TimKDJ commented 3 years ago

I'm assuming you're using MacOS, but are not on Catalina? If so, see if this solves your problem: https://github.com/jasp-stats/jasp-issues/issues/899#issuecomment-662883316

kornbrot commented 3 years ago

I am on Catalina, sorry should have said. Also having trouble with contrasts in nonBayes generalised linear because if select only 2 levels of factor within specify contests still want to know about 3 levels Will have to stick with R or SPSS best Diana

On 1 Sep 2020, at 09:55, Tim de Jong notifications@github.com<mailto:notifications@github.com> wrote:

I'm assuming you're using MacOS, but are not on Catalina? If so, see if this solves your problem: #899 (comment)https://github.com/jasp-stats/jasp-issues/issues/899#issuecomment-662883316

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://github.com/jasp-stats/jasp-issues/issues/957#issuecomment-684605669, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ABF5WKW72TPNDZOLRHJQKUTSDSZHXANCNFSM4QROU6HA.

University of Hertfordshire College Lane, Hatfield, Hertfordshire AL10 9AB, UK +44 (0) 208 444 2081 +44 (0) 7403 18 16 12 d.e.kornbrot@herts.ac.ukmailto:d.e.kornbrot@herts.ac.uk http://dianakornbrot.wordpress.com/ skype: kornbrotme Save our in-boxes! http://emailcharter.org

TimKDJ commented 3 years ago

Could you verify that you clicked the correct installer at https://jasp-stats.org/download/? On Catalina you need to install JASP with http://static.jasp-stats.org/JASP-0.13.1.0-Catalina.dmg On High Sierra and Mojave you need http://static.jasp-stats.org/JASP-0.13.1.0-preCatalina.dmg

I've only seen the error you reported when the MacOS version did not line up with the correct installer.

Also having trouble with contrasts in nonBayes generalised linear because if select only 2 levels of factor within specify contests still want to know about 3 levels

Could you open a separate issue for this?

kornbrot commented 3 years ago

Thanks Done that Now works for Bayes linear mixed, but had been hoping to see Bayes Factor and p(null), p(fixed= within) Have output warning message in red, query in green Bayesian Linear Mixed Models Intercept 95% CI Estimate SE Lower Upper R-hat ESS (bulk) ESS (tail) 0.191600 0.011761 0.168930 0.215215 1.000777 5555.293107 5024.904559 Warning: Random slopes of ‘'wiithin'’ for the random effects grouping factor ‘pno’ removed -- Too few observations to estimate random slopes of ''wiithin'' for random effects grouping factor 'pno'. Note. The Hamiltonian Monte Carlo procedure might be inefficient -- 13 transitions exceeded the maximum tree depth. This can be solved by carefully increasing 'Maximum tree depth Note. The following variable is used as a random effects grouping factor: 'pno'.

wiithin (marginal means) 95% CI Level Estimate SE Lower Upper R-hat ESS (bulk) ESS (tail) 1 0.304221 0.016049 0.273864 0.334919 1.000741 7350.147607 5579.822069 2 0.141967 0.016541 0.110450 0.174133 1.000092 7846.977208 5175.388757 3 0.128612 0.015586 0.098505 0.158689 0.999904 7207.913554 5287.731435 Warning: Random slopes of ‘'wiithin'’ for the random effects grouping factor ‘pno’ removed -- Too few observations to estimate random slopes of ''wiithin'' for random effects grouping factor 'pno'. Note. The Hamiltonian Monte Carlo procedure might be inefficient -- 13 transitions exceeded the maximum tree depth. This can be solved by carefully increasing 'Maximum tree depth Note. The following variable is used as a random effects grouping factor: 'pno'.

Fit Statistics WAIC SE (WAIC) LOO SE (LOO) -303.372167 39.709131 -298.658479 40.068873 Warning: LOO estimate unreliable -- There were 2 observations with the shape parameter (k) of the generalized Pareto distribution higher than > .5. Warning: WAIC estimate unreliable -- There were 31 p_waic estimates larger than 0.4. We recommend using LOO instead.

Fixed Effects Estimates 95% CI Term Estimate SE Lower Upper R-hat ESS Intercept 0.191600 0.011759 0.168930 0.215215 0.999807 5633.465515 wiithin (1) -0.009444 0.013349 -0.036278 0.017029 0.999728 14270.323519 wiithin (2) 0.137932 0.012783 0.112810 0.162874 0.999873 10450.007503 Note. The intercept corresponds to the (unweighted) grand mean; for each factor with k levels, k - 1 parameters are estimated. Consequently, the estimates cannot be directly mapped to factor levels.

Variance/Correlation Estimates pno: Variance Estimates Term Std. Deviation Variance Intercept 0.096796 0.009369 Note. The intercept corresponds to the (unweighted) grand mean; for each factor with k levels, k - 1 parameters are estimated. Consequently, the estimates cannot be directly mapped to factor levels.

Residual Variance Estimates Std. Deviation Variance 0.141574 0.020043

Estimated Marginal Means 95% HPD wiithin Median Lower Upper 1 0.304182 0.275143 0.335994 2 0.141678 0.109507 0.173072 3 0.128587 0.098106 0.158077 But get following for Bayes generalised mixed This analysis terminated unexpectedly.

Error in if (var_name ≠ 'Intercept' && nrow(temp_summary) > 1) {: missing value where TRUE/FALSE needed

Stack trace

To receive assistance with this problem, please report the message above at: https://jasp-stats.org/bug-reports

On 1 Sep 2020, at 10:43, Tim de Jong notifications@github.com<mailto:notifications@github.com> wrote:

Could you verify that you clicked the correct installer at https://jasp-stats.org/download/? On Catalina you need to install JASP with http://static.jasp-stats.org/JASP-0.13.1.0-Catalina.dmg On High Sierra and Mojave you need http://static.jasp-stats.org/JASP-0.13.1.0-preCatalina.dmg

I've only seen the error you reported when the MacOS version did not line up with the correct installer.

Also having trouble with contrasts in nonBayes generalised linear because if select only 2 levels of factor within specify contests still want to know about 3 levels

Could you open a separate issue for this?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://github.com/jasp-stats/jasp-issues/issues/957#issuecomment-684676920, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ABF5WKWKCJQBV2PUCSQXPILSDS635ANCNFSM4QROU6HA.

University of Hertfordshire College Lane, Hatfield, Hertfordshire AL10 9AB, UK +44 (0) 208 444 2081 +44 (0) 7403 18 16 12 d.e.kornbrot@herts.ac.ukmailto:d.e.kornbrot@herts.ac.uk http://dianakornbrot.wordpress.com/ skype: kornbrotme Save our in-boxes! http://emailcharter.org

TimKDJ commented 3 years ago

Have output warning message in red, query in green

I'm sorry, but I'm not sure what you mean by this. When you reply by email to the GitHub issues all attachments and styling is lost. It's much better to go to https://github.com, log into your GitHub account and then post in this issue. You'll also be able to attach the .jasp file (either rename .jasp to .zip, or compress it to .zip directly), because I'm doubtful we'll be able to solve r in if (var_name ≠ 'Intercept' && nrow(temp_summary) > 1) {: missing value where TRUE/FALSE needed without having a copy of your analysis.

kornbrot commented 3 years ago

Ok

On 1 Sep 2020, at 12:03, Tim de Jong notifications@github.com<mailto:notifications@github.com> wrote:

Have output warning message in red, query in green

I'm sorry, but I'm not sure what you mean by this. When you reply by email to the GitHub issues all attachments and styling is lost. It's much better to go to https://github.comhttps://github.com/, log into your GitHub account and then post in this issuehttps://github.com/jasp-stats/jasp-issues/issues/957. You'll also be able to attach the .jasp file (either rename .jasp to .zip, or compress it to .zip directly), because I'm doubtful we'll be able to solve r in if (var_name ≠ 'Intercept' && nrow(temp_summary) > 1) {: missing value where TRUE/FALSE needed without having a copy of your analysis.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://github.com/jasp-stats/jasp-issues/issues/957#issuecomment-684770700, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ABF5WKTQRVZM5OCJYOBLD6DSDTIHLANCNFSM4QROU6HA.

University of Hertfordshire College Lane, Hatfield, Hertfordshire AL10 9AB, UK +44 (0) 208 444 2081 +44 (0) 7403 18 16 12 d.e.kornbrot@herts.ac.ukmailto:d.e.kornbrot@herts.ac.uk http://dianakornbrot.wordpress.com/ skype: kornbrotme Save our in-boxes! http://emailcharter.org

FBartos commented 3 years ago

Hi Diana, thanks for the report.

Now works for Bayes linear mixed, but had been hoping to see Bayes Factor and p(null), p(fixed= within)

Unfortunately, we do not support Bayes factors for mixed effect models yet. We are planning to add them eventually but it won't be in the next release.

Error in if (var_name ≠ 'Intercept' && nrow(temp_summary) > 1) {: missing value where TRUE/FALSE needed

Regarding this error, could you do as Tim suggested and upload the zipped .jasp file than can be produced from JASP?

Cheers, František