Error in nnet

[ddsjoberg]

Hey hey,

This issue was reported today for a multinomial model. I didn't get a chance to look into the details yet, but wanted to record it here.

library(nnet)
library(MASS)
library(magrittr)

test1 <- multinom(race ~ age + lwt + bwt, data = birthwt)
#> # weights:  15 (8 variable)
#> initial  value 207.637723 
#> iter  10 value 168.608572
#> final  value 168.587348 
#> converged

broom::tidy(test1) %>% knitr::kable()

y.level	term	estimate	std.error	statistic	p.value
2	(Intercept)	2.0502611	0.0009135	2244.476995	0.0000000
2	age	-0.1464569	0.0394259	-3.714736	0.0002034
2	lwt	0.0205184	0.0068180	3.009432	0.0026174
2	bwt	-0.0009939	0.0002851	-3.485562	0.0004911
3	(Intercept)	3.9546842	0.0004048	9769.456437	0.0000000
3	age	-0.0553396	0.0285068	-1.941279	0.0522245
3	lwt	-0.0129022	0.0059785	-2.158092	0.0309207
3	bwt	-0.0004768	0.0002111	-2.258674	0.0239037

broom.helpers::tidy_plus_plus(test1)
#> Error: Join columns must be present in data.
#> x Problem with `term`.

^{Created on 2021-02-24 by the reprex package (v1.0.0)}

[larmarange]

Thanks for the feedback. It seems related to add_n feature.

library(nnet)
library(MASS)
library(broom.helpers)

test1 <- multinom(race ~ age + lwt + bwt, data = birthwt)
#> # weights:  15 (8 variable)
#> initial  value 207.637723 
#> iter  10 value 168.608572
#> final  value 168.587348 
#> converged
test1 %>% tidy_plus_plus()
#> Error: Join columns must be present in data.
#> x Problem with `term`.
test1 %>% tidy_plus_plus(add_n = FALSE)
#> # A tibble: 6 x 17
#>   y.level term  variable var_label var_class var_type var_nlevels contrasts
#>   <chr>   <chr> <chr>    <chr>     <chr>     <chr>          <int> <chr>    
#> 1 2       age   age      age       integer   continu~          NA <NA>     
#> 2 2       lwt   lwt      lwt       integer   continu~          NA <NA>     
#> 3 2       bwt   bwt      bwt       integer   continu~          NA <NA>     
#> 4 3       age   age      age       integer   continu~          NA <NA>     
#> 5 3       lwt   lwt      lwt       integer   continu~          NA <NA>     
#> 6 3       bwt   bwt      bwt       integer   continu~          NA <NA>     
#> # ... with 9 more variables: contrasts_type <chr>, reference_row <lgl>,
#> #   label <chr>, estimate <dbl>, std.error <dbl>, statistic <dbl>,
#> #   p.value <dbl>, conf.low <dbl>, conf.high <dbl>
test1 %>% tidy_and_attach() %>% tidy_add_n()
#> Error: Join columns must be present in data.
#> x Problem with `term`.

^{Created on 2021-02-24 by the reprex package (v1.0.0)}

[larmarange]

OK the problem comes from the fact that race is not coded as a factor. Easy solution: transform race before computing the model

library(nnet)
library(MASS)
library(broom.helpers)

tmp <- birthwt
tmp$race <- factor(tmp$race)
test2 <- multinom(race ~ age + lwt + bwt, data = tmp)
#> # weights:  15 (8 variable)
#> initial  value 207.637723 
#> iter  10 value 168.608572
#> final  value 168.587348 
#> converged
test2 %>% tidy_plus_plus()
#> # A tibble: 6 x 19
#>   y.level term  variable var_label var_class var_type var_nlevels contrasts
#>   <chr>   <chr> <chr>    <chr>     <chr>     <chr>          <int> <chr>    
#> 1 2       age   age      age       integer   continu~          NA <NA>     
#> 2 2       lwt   lwt      lwt       integer   continu~          NA <NA>     
#> 3 2       bwt   bwt      bwt       integer   continu~          NA <NA>     
#> 4 3       age   age      age       integer   continu~          NA <NA>     
#> 5 3       lwt   lwt      lwt       integer   continu~          NA <NA>     
#> 6 3       bwt   bwt      bwt       integer   continu~          NA <NA>     
#> # ... with 11 more variables: contrasts_type <chr>, reference_row <lgl>,
#> #   label <chr>, n_obs <dbl>, n_event <dbl>, estimate <dbl>, std.error <dbl>,
#> #   statistic <dbl>, p.value <dbl>, conf.low <dbl>, conf.high <dbl>

^{Created on 2021-02-24 by the reprex package (v1.0.0)}

[larmarange]

Should be fixed now

library(nnet)
library(MASS)
library(broom.helpers)

test1 <- multinom(race ~ age + lwt + bwt, data = birthwt)
#> # weights:  15 (8 variable)
#> initial  value 207.637723 
#> iter  10 value 168.608572
#> final  value 168.587348 
#> converged
test1 %>% tidy_plus_plus()
#> # A tibble: 6 x 19
#>   y.level term  variable var_label var_class var_type var_nlevels contrasts
#>   <chr>   <chr> <chr>    <chr>     <chr>     <chr>          <int> <chr>    
#> 1 2       age   age      age       integer   continu~          NA <NA>     
#> 2 2       lwt   lwt      lwt       integer   continu~          NA <NA>     
#> 3 2       bwt   bwt      bwt       integer   continu~          NA <NA>     
#> 4 3       age   age      age       integer   continu~          NA <NA>     
#> 5 3       lwt   lwt      lwt       integer   continu~          NA <NA>     
#> 6 3       bwt   bwt      bwt       integer   continu~          NA <NA>     
#> # ... with 11 more variables: contrasts_type <chr>, reference_row <lgl>,
#> #   label <chr>, n_obs <dbl>, n_event <dbl>, estimate <dbl>, std.error <dbl>,
#> #   statistic <dbl>, p.value <dbl>, conf.low <dbl>, conf.high <dbl>

^{Created on 2021-02-24 by the reprex package (v1.0.0)}

[ddsjoberg]

Wowow, so fast! Thanks