Search code examples
rstatisticsformulalogical-operators

Strange behaviour of == with formula

I am a bit puzzled by the following. I have two formulas and would like to check whether they are the same. Here I expect to get FALSE returned.

fm1 <- formula(schades ~ termijn + zipcode + provincie + regionvormgemeente + energielabel + trede)
fm2 <- formula(schades ~ termijn + zipcode + provincie + regionvormgemeente + energielabel)
fm1 == fm2
#> [1] TRUE

identical(fm1, fm2)
#> [1] FALSE

What is the reason that fm1 == fm2 returns TRUE?

Created on 2021-12-17 by the reprex package (v2.0.1)

Solution

  • == is designed to compare values in atomic vectors, not formulars.

    Furthermore, see the following example from ?== :

    x1 <- 0.5 - 0.3
x2 <- 0.3 - 0.1
x1 == x2                   # FALSE on most machines
isTRUE(all.equal(x1, x2))  # TRUE everywhere

    Applied to your example you can find:

        > fm1 <- formula(schades ~ termijn + zipcode + provincie + regionvormgemeente + energielabel + trede)
> fm2 <- formula(schades ~ termijn + zipcode + provincie + regionvormgemeente + energielabel)
> fm1 == fm2
[1] TRUE
> 
> all.equal(fm1, fm2)
[1] "formulas differ in contents"
> isTRUE(all.equal(fm1,fm2))
[1] FALSE

    But apparently reducing the number of predictors returns the expected result. It just illustrates that == should not be used for this type of comparison as its behaviour is not coherent:

    > fm1 <- formula(schades ~ termijn + zipcode + provincie)
> fm2 <- formula(schades ~ termijn + zipcode)
> fm1 == fm2
[1] FALSE
> isTRUE(all.equal(fm1,fm2))
[1] FALSE