Minus sign in FixedEffectModels not handled

[eloualiche]

This might an edge case coming from FixedEffectModels permissive formulas.
Fixedeffectmodels allows for some "calculations" inside the formulas, one of them being a substraction. This returns a valid result in fixedeffectmodels but triggers an error once inside of regtable

using DataFrames
using RDatasets
using RegressionTables
using FixedEffectModels
iris = dataset("datasets", "iris")
r1 = reg(iris, @formula(SepalLength  ~ SepalWidth - PetalLength ))
regtable(r1; render = AsciiTable())

The output is

julia> r1 = reg(iris, @formula(SepalLength  ~ SepalWidth - PetalLength ))
                                    FixedEffectModel
=========================================================================================
Number of obs:                          150   Converged:                             true
dof (model):                              1   dof (residuals):                        147
R²:                                   0.638   R² adjusted:                          0.635
F-statistic:                        260.425   P-value:                              0.000
=========================================================================================
                           Estimate  Std. Error    t-stat  Pr(>|t|)  Lower 95%  Upper 95%
─────────────────────────────────────────────────────────────────────────────────────────
SepalWidth - PetalLength  -0.332045   0.0205758  -16.1377    <1e-33  -0.372708  -0.291383
(Intercept)                5.61068    0.0433073  129.555     <1e-99   5.52509    5.69627
=========================================================================================


julia> regtable(r1; render = AsciiTable())
ERROR: MethodError: no method matching iterate(::FunctionTerm{typeof(-), Vector{Term}})
...

I am not sure it's worth doing anything about. Maybe people can find this issue if they run into something similar.
I do believe this is new behavior that showed up as I was running some old code (could not tell you the version though).

[junder873]

Thank you for bringing this to my attention.

Looked into this, turns out the issue isn't necessarily the subtraction term but the singe right hand side term. For example:

r1 = reg(iris, @formula(SepalLength ~ log1p(SepalWidth))
regtable(r1)

Also causes an error.

I am working on a fix, but if you need a quick fix you can add the intercept directly which will solve the issue:

r1 = reg(iris, @formula(SepalLength ~ 1 + (SepalWidth - PetalLength))
regtable(r1)
--------------------------------------
                           SepalLength
--------------------------------------
(Intercept)                   5.611***
                               (0.043)
SepalWidth - PetalLength     -0.332***
                               (0.021)
--------------------------------------
N                                  150
R2                               0.638
--------------------------------------