jtools
jtools copied to clipboard
jacob-long
Cannot exponentiate coefficients of MASS:polr object with export_summs
Brief description of the problem
Cannot exponentiate the coefficients of MASS::polr-object (whether after multiple impuation or not) with export_summs function
I've asked this on Stackoverflow but later thought that this could potentially be a bug. Here I have extended the code to make debugging easier.
library(mice)
#>
#> Attaching package: 'mice'
#> The following object is masked from 'package:stats':
#>
#> filter
#> The following objects are masked from 'package:base':
#>
#> cbind, rbind
library(MASS)
library(jtools)
library(reprex)
set.seed(3)
# creating a dataset and imputing it
y=sample(c(1:4,NA),size=50, replace=T)
y=as.ordered(x)
#> Error in is.ordered(x): object 'x' not found
x=sample(c(1:4,NA),size=50, replace=T)
z=sample(c(1,2),size=50, replace=T)
q <- runif(50)
d <- data.frame(cbind(x,y,z,q))
mi_d <- mice(d,m=2,maxit=2)
#>
#> iter imp variable
#> 1 1 x y
#> 1 2 x y
#> 2 1 x y
#> 2 2 x y
# ordinal regression
polr_model <- polr(as.factor(y) ~ x , Hess=TRUE)
mi_polr <- with(mi_d,polr(as.factor(y) ~ x , Hess=TRUE))
POOLED_P <- pool(mi_polr)
# exponientiating coefficients works fine when plotting
plot_summs(POOLED_P,exp=T)
#> Loading required namespace: broom.mixed
#.. but not with exporting the table
jtools::export_summs(POOLED_P,exp=T)
#> Warning in nobs.default(m, use.fallback = TRUE): no 'nobs' method is available
#> Warning in knit_print.huxtable(x, ...): Unrecognized output format "gfm". Using `to_screen` to print huxtables.
#> Set options("huxtable.knitr_output_format") manually to "latex", "html", "rtf", "docx", "pptx", "md" or "screen".
─────────────────────────────────────────────────
Model 1
─────────────────────────
x 0.49
(0.40)
1|2 0.12
(0.93)
2|3 1.48
(0.94)
3|4 2.28 *
(0.89)
─────────────────────────
nobs 0
nimp 2.00
nobs.1 50.00
─────────────────────────────────────────────────
*** p < 0.001; ** p < 0.01; * p < 0.05.
Column names: names, Model 1
jtools::export_summs(polr_model, exp=T)
#> Warning: The `x` argument of `as_tibble.matrix()` must have unique column names if `.name_repair` is omitted as of tibble 2.0.0.
#> Using compatibility `.name_repair`.
#> Warning in FUN(X[[i]], ...): tidy() does not return p values for models of class
#> data.frame; significance stars not printed.
#> Warning in knit_print.huxtable(x, ...): Unrecognized output format "gfm". Using `to_screen` to print huxtables.
#> Set options("huxtable.knitr_output_format") manually to "latex", "html", "rtf", "docx", "pptx", "md" or "screen".
─────────────────────────────────────────────────
Model 1
─────────────────────────
x 0.39
(0.34)
1|2 -0.27
(0.86)
2|3 0.97
(0.86)
3|4 1.93
(0.94)
─────────────────────────
nobs 29.00
edf 4.00
logLik -39.32
AIC 86.65
BIC 92.12
deviance 78.65
df.residual 25.00
nobs.1 29.00
─────────────────────────────────────────────────
*** p < 0.001; ** p < 0.01; * p < 0.05.
Column names: names, Model 1
# let's try with regular glm
glm_model <- glm(q~z,data=d,family=binomial())
#> Warning in eval(family$initialize): non-integer #successes in a binomial glm!
jtools::export_summs(glm_model,exp=F)
#> Warning in eval(family$initialize): non-integer #successes in a binomial glm!
#> Warning in eval(family$initialize): non-integer #successes in a binomial glm!
#> Warning in knit_print.huxtable(x, ...): Unrecognized output format "gfm". Using `to_screen` to print huxtables.
#> Set options("huxtable.knitr_output_format") manually to "latex", "html", "rtf", "docx", "pptx", "md" or "screen".
─────────────────────────────────────────────────
Model 1
─────────────────────────
(Intercept) 0.07
(0.93)
z -0.04
(0.57)
─────────────────────────
N 50
AIC 73.35
BIC 77.17
Pseudo R2 -0.00
─────────────────────────────────────────────────
*** p < 0.001; ** p < 0.01; * p < 0.05.
Column names: names, Model 1
jtools::export_summs(glm_model, exp=T) # recognizes model1 as a glm object and exponentiates the coefficients
#> Warning in eval(family$initialize): non-integer #successes in a binomial glm!
#> Warning in eval(family$initialize): Unrecognized output format "gfm". Using `to_screen` to print huxtables.
#> Set options("huxtable.knitr_output_format") manually to "latex", "html", "rtf", "docx", "pptx", "md" or "screen".
─────────────────────────────────────────────────
Model 1
─────────────────────────
(Intercept) 1.08
(0.93)
z 0.96
(0.57)
─────────────────────────
N 50
AIC 73.35
BIC 77.17
Pseudo R2 -0.00
─────────────────────────────────────────────────
*** p < 0.001; ** p < 0.01; * p < 0.05.
Column names: names, Model 1
Created on 2021-05-07 by the reprex package (v2.0.0)
I'll take a look at whether I can do an interim solution here although there's a chance I can't deal with this without making polr models a "first class" supported model in the package, which is something I plan to do (#16)
