`standardize()` fails in some cases

[etiennebacher]

Cf easystats/report#379

library(datawizard)

data(sleep)
d <- data_modify(sleep, group = as.integer(group) - 1L)
d_wide <- data_to_wide(
  d,
  names_from = "group",
  values_from = "extra",
  names_prefix = "group"
)

model1 <- lm(d_wide$group0 - d_wide$group1 ~ 1)
standardize(model1)
#> Warning: Using `$` in model formulas can produce unexpected results. Specify your
#>   model using the `data` argument instead.
#>   Try: group0 - group1 ~ 1,
#>   data =

#> Warning: Using `$` in model formulas can produce unexpected results. Specify your
#>   model using the `data` argument instead.
#>   Try: group0 - group1 ~ 1,
#>   data =

#> Warning: Using `$` in model formulas can produce unexpected results. Specify your
#>   model using the `data` argument instead.
#>   Try: group0 - group1 ~ 1,
#>   data =
#> 
#> Call:
#> lm(formula = d_wide$group0 - d_wide$group1 ~ 1, data = data_std)
#> 
#> Coefficients:
#> (Intercept)  
#>       -1.58

model1 <- lm(group0 - group1 ~ 1, data = d_wide)
standardize(model1)
#> Error in eval(predvars, data, env): objet 'group0' introuvable
#> Error: Unable to refit the model with standardized data.
#>   Try instead to standardize the data (standardize(data)) and refit the
#>   model manually.

^{Created on 2023-07-01 with reprex v2.0.2}

[etiennebacher]

Looks like it comes from these lines:

datawizard/R/standardize.models.R

Lines 81 to 89 in 3ff2209

	data_std <- NULL # needed to avoid note
	.standardize_models(x,
	robust = robust, two_sd = two_sd,
	weights = weights,
	verbose = verbose,
	include_response = include_response,
	update_expr = stats::update(x, data = data_std),
	...
	)

stats::update() fails in this case when data = NULL:

library(datawizard)

data(sleep)
d <- data_modify(sleep, group = as.integer(group) - 1L)
d_wide <- data_to_wide(
  d,
  names_from = "group",
  values_from = "extra",
  names_prefix = "group"
)

model1 <- lm(d_wide$group0 - d_wide$group1 ~ 1)
stats::update(model1, data = NULL)
#> 
#> Call:
#> lm(formula = d_wide$group0 - d_wide$group1 ~ 1, data = NULL)
#> 
#> Coefficients:
#> (Intercept)  
#>       -1.58

model2 <- lm(group0 - group1 ~ 1, data = d_wide)
stats::update(model2, data = NULL)
#> Error in eval(predvars, data, env): objet 'group0' introuvable

@strengejacke I think you wrote this code, do you need how to fix this?

[strengejacke]

Minimal reprex:

data(iris)
model <- lm(Sepal.Length ~ 1, data = iris)
stats::update(model, data = NULL)
#> Error in eval(predvars, data, env): object 'Sepal.Length' not found

^{Created on 2023-07-01 with reprex v2.0.2}

[strengejacke]

The problem for your particular example is the returned data:

library(datawizard)

data(sleep)
d <- data_modify(sleep, group = as.integer(group) - 1L)
d_wide <- data_to_wide(
  d,
  names_from = "group",
  values_from = "extra",
  names_prefix = "group"
)

model2 <- lm(group0 - group1 ~ 1, data = d_wide)
head(insight::get_data(model2, source = "mf"))
#>   group0 - group1
#> 1            -1.2
#> 2            -2.4
#> 3            -1.3
#> 4            -1.3
#> 5             0.0
#> 6            -1.0

we need the variables group0 and group1.

The PR at least works for these cases:

library(datawizard)

data(iris)
model <- lm(Sepal.Length ~ 1, data = iris)
standardize(model)
#> 
#> Call:
#> lm(formula = Sepal.Length ~ 1, data = data_std)
#> 
#> Coefficients:
#> (Intercept)  
#>  -4.079e-16

data(sleep)
d <- data_modify(sleep, group = as.integer(group) - 1L)
d_wide <- data_to_wide(
  d,
  names_from = "group",
  values_from = "extra",
  names_prefix = "group"
)

model1 <- lm(d_wide$group0 - d_wide$group1 ~ 1)
standardize(model1)
#> Warning: Using `$` in model formulas can produce unexpected results. Specify your
#>   model using the `data` argument instead.
#>   Try: group0 - group1 ~ 1,
#>   data =

#> Warning: Using `$` in model formulas can produce unexpected results. Specify your
#>   model using the `data` argument instead.
#>   Try: group0 - group1 ~ 1,
#>   data =

#> Warning: Using `$` in model formulas can produce unexpected results. Specify your
#>   model using the `data` argument instead.
#>   Try: group0 - group1 ~ 1,
#>   data =

#> Warning: Using `$` in model formulas can produce unexpected results. Specify your
#>   model using the `data` argument instead.
#>   Try: group0 - group1 ~ 1,
#>   data =

#> Warning: Using `$` in model formulas can produce unexpected results. Specify your
#>   model using the `data` argument instead.
#>   Try: group0 - group1 ~ 1,
#>   data =
#> 
#> Call:
#> lm(formula = d_wide$group0 - d_wide$group1 ~ 1, data = data_std)
#> 
#> Coefficients:
#> (Intercept)  
#>       -1.58

model2 <- lm(group0 - group1 ~ 1, data = d_wide)
standardize(model2)
#> 
#> Call:
#> lm(formula = group0 - group1 ~ 1, data = data_std)
#> 
#> Coefficients:
#> (Intercept)  
#>   5.266e-17

^{Created on 2023-07-01 with reprex v2.0.2}

[mattansb]

Note that if you have a composite variable (as in this example), and you standardize each part separately, you can easily get nonsense results...

$$ Z(x) - Z(z) \neq Z(x - z) $$

[LAhmos]

any workaround this?

[mattansb]

You would probably want to compute the composite variable beforehand (not in the formula):

library(datawizard)

data(sleep)

d_wide <- sleep |> 
  data_modify(group = as.integer(group) - 1L) |> 
  data_to_wide(
    names_from = "group",
    values_from = "extra",
    names_prefix = "group"
  ) |> 
  data_modify(
    diff = group0 - group1
  )

model1 <- lm(diff ~ 1, data = d_wide)
standardize(model1)
#> 
#> Call:
#> lm(formula = diff ~ 1, data = data_std)
#> 
#> Coefficients:
#> (Intercept)  
#>   1.755e-17 
#>   

[drpradeepharish]

@strengejacke

Does the error still persist?

library(tidyverse)    
library(easystats)

    
    Data <- structure(list(`12_month_remission` = c(0, 1, 1, 0, 0, 0, 0, 
    0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 1, 0, 0, 0, 0, 1, 0, 
    1, 1, 0, 1, 1, 0, 0, 0, 1, 0, 1, 0, 1, 1, 0, 1, 0, 0, 0, 0, 0, 
    1), Sex = structure(c(2L, 1L, 1L, 2L, 2L, 2L, 1L, 2L, 2L, 1L, 
    1L, 1L, 2L, 1L, 2L, 1L, 2L, 2L, 2L, 1L, 2L, 1L, 1L, 1L, 2L, 2L, 
    1L, 1L, 2L, 1L, 1L, 1L, 1L, 2L, 1L, 2L, 1L, 2L, 1L, 1L, 2L, 1L, 
    2L, 1L, 2L, 1L, 2L, 1L, 2L, 1L), levels = c("Female", "Male"), class = "factor")), row.names = c(NA, 
    -50L), class = c("tbl_df", "tbl", "data.frame"))

glm(`12_month_remission` ~ Sex, family = "binomial", data = Data) %>% 
  report::report() 

Error in eval(predvars, data, env) : object '12_month_remission' not found
Error: Unable to refit the model with standardized data.
Try instead to standardize the data (standardize(data)) and refit the model manually.