ggpredict unable to plot from a natural spline with specified knots #240

ifeomaozo · 2021-10-23T20:42:28Z

@strengejacke, I am reopening this issue since the insight update did not solve the problem. Please see below

> library(glmmTMB)
> library(splines)
> 
> data_e <- data.frame(
+   positives = c(3L, 2L, 0L, 7L, 0L, 5L, 0L, 3L, 6L, 0L, 2L, 0L, 2L, 3L, 2L, 14L, 1L, 3L, 3L, 0L), 
+   values = c(86.1702127659574, 100, 81.0344827586207, 98.7341772151899, 97.5903614457831, 
+              98.1308411214953, 96.551724137931, 100, 100, 96.045197740113, 99.1150442477876, 
+              98.8888888888889, 100, 91.6279069767442, 33.3333333333333, 0, 
+              92.3076923076923, 24, 0, 93.5483870967742))
> 
> m1 <- glmmTMB(
+   positives ~ ns(values, 3, knots = seq(min(values), max(values), length = 4)[2:3]),
+   data = data_e,
+   ziformula =  ~ 1,
+   family = poisson
+ )
> ggeffects::ggpredict(m1, terms="values [all]", type ='zero_inflated')

# Predicted counts of positives
# x = values

Error in cat(insight::format_table(dd, digits = digits, protect_integers = TRUE)) : 
  argument 1 (type 'list') cannot be handled by 'cat'

The text was updated successfully, but these errors were encountered:

strengejacke · 2021-10-23T22:35:08Z

strange, maybe you updated at a time where the new binaries weren't properly built? For me, it works:

library(glmmTMB)
library(splines)

data_e <- data.frame(
  positives = c(3L, 2L, 0L, 7L, 0L, 5L, 0L, 3L, 6L, 0L, 2L, 0L, 2L, 3L, 2L, 14L, 1L, 3L, 3L, 0L), 
  values = c(86.1702127659574, 100, 81.0344827586207, 98.7341772151899, 97.5903614457831, 
             98.1308411214953, 96.551724137931, 100, 100, 96.045197740113, 99.1150442477876, 
             98.8888888888889, 100, 91.6279069767442, 33.3333333333333, 0, 
             92.3076923076923, 24, 0, 93.5483870967742))
m1 <- glmmTMB(
  positives ~ ns(values, 3, knots = seq(min(values), max(values), length = 4)[2:3]),
  data = data_e,
  ziformula =  ~ 1,
  family = poisson
)

ggeffects::ggpredict(m1, terms="values [all]", type ='zero_inflated')
#> # Predicted counts of positives
#> 
#> values | Predicted |        95% CI
#> ----------------------------------
#>   0.00 |      6.41 | [2.65, 10.18]
#>  33.33 |      1.52 | [0.00,  3.04]
#>  86.17 |      1.33 | [0.00,  2.65]
#>  92.31 |      1.82 | [0.52,  3.11]
#>  93.55 |      1.94 | [0.73,  3.15]
#>  96.55 |      2.29 | [1.19,  3.39]
#>  98.13 |      2.50 | [1.36,  3.65]
#> 100.00 |      2.78 | [1.37,  4.18]

^{Created on 2021-10-24 by the reprex package (v2.0.1)}

ifeomaozo · 2021-10-23T23:05:27Z

hmmm...I just updated 2 hours ago. Do you have any troubleshooting suggestions? I uninstalled the previous version of insight on my windows computer and tried to reinstall and got the same error. I have also tried to replicate the problem on a mac but I can't install the updated package. My session info on the mac and windows is below. I will really appreciate any help.

install.packages("insight", repos = "https://easystats.r-universe.dev")
Warning in install.packages :
  unable to access index for repository https://easystats.r-universe.dev/src/contrib:
  cannot open URL 'https://easystats.r-universe.dev/src/contrib/PACKAGES'
Warning in install.packages :
  package ‘insight’ is not available for this version of R

A version of this package for your version of R might be available elsewhere,
see the ideas at
https://cran.r-project.org/doc/manuals/r-patched/R-admin.html#Installing-packages
Warning in install.packages :
  unable to access index for repository https://easystats.r-universe.dev/bin/macosx/contrib/4.1:
  cannot open URL 'https://easystats.r-universe.dev/bin/macosx/contrib/4.1/PACKAGES'
>

sessionInfo(package = NULL)
R version 4.1.1 (2021-08-10)
Platform: x86_64-apple-darwin17.0 (64-bit)
Running under: macOS Mojave 10.14.6

Matrix products: default
BLAS:   /System/Library/Frameworks/Accelerate.framework/Versions/A/Frameworks/vecLib.framework/Versions/A/libBLAS.dylib
LAPACK: /Library/Frameworks/R.framework/Versions/4.1/Resources/lib/libRlapack.dylib

locale:
[1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8

attached base packages:
[1] splines   stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] glmmTMB_1.1.2.3

loaded via a namespace (and not attached):
 [1] Rcpp_1.0.7          lattice_0.20-44     MASS_7.3-54         grid_4.1.1          nlme_3.1-152        minqa_1.2.4        
 [7] nloptr_1.2.2.2      Matrix_1.3-4        boot_1.3-28         ggeffects_1.1.1     lme4_1.1-27.1       tools_4.1.1        
[13] TMB_1.7.22          numDeriv_2016.8-1.1 compiler_4.1.1      insight_0.14.5

sessionInfo(package = NULL)
R version 4.0.2 (2020-06-22)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows 10 x64 (build 17134)

Matrix products: default

locale:
[1] LC_COLLATE=English_United States.1252  LC_CTYPE=English_United States.1252    LC_MONETARY=English_United States.1252
[4] LC_NUMERIC=C                           LC_TIME=English_United States.1252    

attached base packages:
[1] splines   stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] glmmTMB_1.1.2.3

loaded via a namespace (and not attached):
 [1] Rcpp_1.0.7          nloptr_1.2.2.2      pillar_1.4.6        compiler_4.0.2      TMB_1.7.22          class_7.3-17       
 [7] tools_4.0.2         boot_1.3-25         statmod_1.4.34      lme4_1.1-23         digest_0.6.27       packrat_0.6.0      
[13] nlme_3.1-148        evaluate_0.14       lifecycle_1.0.0     tibble_3.0.3        lattice_0.20-41     pkgconfig_2.0.3    
[19] rlang_0.4.10        Matrix_1.2-18       DBI_1.1.0           yaml_2.2.1          xfun_0.21           e1071_1.7-3        
[25] dplyr_1.0.0         knitr_1.29          generics_0.0.2      vctrs_0.3.6         ggeffects_0.16.0    classInt_0.4-3     
[31] sjlabelled_1.1.7    grid_4.0.2          tidyselect_1.1.0    glue_1.4.2          sf_0.9-5            R6_2.4.1           
[37] rmarkdown_2.5       minqa_1.2.4         purrr_0.3.4         magrittr_2.0.1      ellipsis_0.3.1      htmltools_0.5.1.1  
[43] units_0.6-7         MASS_7.3-51.6       insight_0.14.5.1    numDeriv_2016.8-1.1 KernSmooth_2.23-17  crayon_1.4.1

ifeomaozo · 2021-10-24T04:36:41Z

I was able to solve the issue by updating my R version and reinstalling all packages. The problem I am having now is that when I fit my final model, I end up with a new error message. See below. The data can be downloaded here. Again, I will appreciate any help.

m2 <- glmmTMB(positives~ns(edu_a, 3)+ ns(wealth, 3)+ns(pop_density_0m, 2) + ns(median_age, 2)+
                ns(med_treat_fever, knots = seq(min(med_treat_fever),max(med_treat_fever),length =4)[2:3])+
                ns(precipitation_monthly_0m, 3) + ns(EVI_0m, 3)+
                + offset(log(child_6_59_tested_malaria)) +
                mat(pos + 0 | ID) + ar1(month_year + 0 | ID2), data=map2,  ziformula=~1,family=poisson)

ggeffects::ggpredict(m2,terms="edu_a[all]", type ='zero_inflated')
Error in get_predictions_glmmTMB(model, data_grid, ci.lvl, linv, type,  : 
  'list' object cannot be coerced to type 'double'

strengejacke · 2021-11-07T10:19:28Z

Could you also please upload the fitted model object? It takes hours to fit the model on my machine ;-)

ifeomaozo · 2021-11-07T11:55:36Z

@strengejacke, thanks for returning to this! The model rds file can be downloaded here

strengejacke · 2021-11-07T12:16:14Z

Not sure what the issue here exactly is, but the function fails when computing the model matrix for the count component of the model:

library(splines)
library(glmmTMB)
library(ggeffects)

map2 <- readr::read_csv("D:/Downloads/all_data.csv", show_col_types = FALSE)
#> New names:
#> * `` -> ...1
m2 <- readRDS("D:/Downloads/multivariate_model.rds")
nd <- new_data(m2, "edu_a [all]")

condformula <- lme4::nobars(stats::formula(m2)[-2])
stats::model.matrix(condformula, nd)
#> Error in qr.default(t(const)): NA/NaN/Inf in foreign function call (arg 1)

^{Created on 2021-11-07 by the reprex package (v2.0.1)}

nicholasgrebe · 2023-05-04T15:50:41Z

I am having the same issue as @ifeomaozo -- I get this error when trying to use the ggpredict() call and "zero_inflated" type on a fitted glmmTMB object:

"Error in get_predictions_glmmTMB(model, data_grid, ci.lvl, linv, type, :
'list' object cannot be coerced to type 'double'"

Just moving this to the front in case there's been some resolution on it. Thank you!

jml-science · 2023-10-06T13:53:22Z

@nicholasgrebe Could you resolve the issue by any chance? I am trying to do the same thing (ggpredict with "zero_inflated" type on a fitted glmmTMB object) and receive the same error. Curiously, it just happens with one model specification - for other models, the same command works...

nicholasgrebe · 2023-10-06T14:22:12Z

@jml-science I never had any luck with this package, but using the marginaleffects package was a good substitute (e.g. the "type" argument under the plot_predictions function did what I needed): https://marginaleffects.com/articles/reference/plot_predictions.html

strengejacke · 2023-10-06T20:29:45Z

Note that this issue (#240 (comment)) stems from how standard errors for zero-inflated models are calculated (or better: should be calculated, see https://journal.r-project.org/archive/2017/RJ-2017-066/index.html), so I'm not sure how accurate SE's returned by marginaleffects or predict() are. ggemmeans() still may work, though.

library(ggeffects)
library(marginaleffects)
library(glmmTMB)
data(Salamanders)
m1 <- glmmTMB(count ~ mined + (1 | site),
  zi = ~mined,
  family = poisson, data = Salamanders
)
ggpredict(m1, terms = "mined", type = "zero_inflated")
#> # Predicted counts of count
#> 
#> mined | Predicted |       95% CI
#> --------------------------------
#> yes   |      0.26 | [0.10, 0.43]
#> no    |      2.21 | [1.77, 2.64]
#> 
#> Adjusted for:
#> * site = NA (population-level)
avg_predictions(m1, variables = "mined", type = "response", vcov = vcov(m1))
#> 
#>  mined Estimate Std. Error     z Pr(>|z|)     S 2.5 % 97.5 %
#>    yes    0.272     0.0442  6.15   <0.001  30.2 0.185  0.359
#>    no     2.265     0.1583 14.31   <0.001 151.8 1.955  2.576
#> 
#> Columns: mined, estimate, std.error, statistic, p.value, s.value, conf.low, conf.high 
#> Type:  response

^{Created on 2023-10-06 with reprex v2.0.2}

Another issue is that there were some changes in glmmTMB that changed the behaviour of predict for model with truncated family, see easystats/insight#679 (comment). Maybe when I try to fix #319, this issue here can also be resolved

strengejacke added 3 investigators ❔❓ bug 🐛 Something isn't working labels Nov 7, 2021

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

ggpredict unable to plot from a natural spline with specified knots #240

ggpredict unable to plot from a natural spline with specified knots #240

ifeomaozo commented Oct 23, 2021

strengejacke commented Oct 23, 2021

ifeomaozo commented Oct 23, 2021 •

edited

Loading

ifeomaozo commented Oct 24, 2021 •

edited by strengejacke

Loading

strengejacke commented Nov 7, 2021

ifeomaozo commented Nov 7, 2021 •

edited by strengejacke

Loading

strengejacke commented Nov 7, 2021

nicholasgrebe commented May 4, 2023

jml-science commented Oct 6, 2023

nicholasgrebe commented Oct 6, 2023

strengejacke commented Oct 6, 2023

ggpredict unable to plot from a natural spline with specified knots #240

ggpredict unable to plot from a natural spline with specified knots #240

Comments

ifeomaozo commented Oct 23, 2021

strengejacke commented Oct 23, 2021

ifeomaozo commented Oct 23, 2021 • edited Loading

ifeomaozo commented Oct 24, 2021 • edited by strengejacke Loading

strengejacke commented Nov 7, 2021

ifeomaozo commented Nov 7, 2021 • edited by strengejacke Loading

strengejacke commented Nov 7, 2021

nicholasgrebe commented May 4, 2023

jml-science commented Oct 6, 2023

nicholasgrebe commented Oct 6, 2023

strengejacke commented Oct 6, 2023

ifeomaozo commented Oct 23, 2021 •

edited

Loading

ifeomaozo commented Oct 24, 2021 •

edited by strengejacke

Loading

ifeomaozo commented Nov 7, 2021 •

edited by strengejacke

Loading