report_table() for family = "quasibinomial"

[bwiernik]

I'm also encountering a similar problem. I installed easystats collection after installing report individually. I tried reinstalling report separately, but no luck so far.

The issue appears to be situational. I can run a simple linear model with no problems using report(), but with a glm I get the error (see below). The response variable data is between 0 and 1 (eg. 0.00, 0.102, ... .899, 1.00), and the explanatory variables are categorical. Running a similar linear model, but with a different response variables works fine. I suspect it's something with the GLM() and the response variable data type. I'll continue testing to see if I figure out the root of the issue.

glm( formula = "Alpha.Score ~ Diet",
                      data = subset(dt.alphaPlus.T0.melt, Alpha.Metric == "Observed"),
                      family = "quasibinomial") %>% report::report()

Error in UseMethod("format_value") : 
no applicable method for 'format_value' applied to an object of class "NULL"

Originally posted by @sielerjm in #56 (comment)

[bwiernik]

Issue is this line in report_table.glm:

if (insight::model_info(x)$is_logit) {
    params <- data_remove(params, "df_error")
  }

family = "quasibinomial" are logit, but also have a t (not z) distribution.

It should be

mi <- insight::model_info(x)
if (mi$is_logit && mi$family != "quasibinomial") {
    params <- data_remove(params, "df_error")
  }

or

mi <- insight::model_info(x)
if (mi$is_binomial && mi$family != "quasibinomial") {
    params <- data_remove(params, "df_error")
  }

[sielerjm]

Based on @bwiernik comment, I reran the tests and it appears the issue is with "quasibinomial" family (see image).

[John-Mola]

Anyone ever come up with a solution to this?

[sielerjm]

Anyone ever come up with a solution to this?

Still no resolution, afaik.