Convert change in probability to probability / odds

[mattansb]

Closes #660

• NEWS / Version bump
  • we also dropped log(rr)
• Make sure all oddsratio_to_<change>() functions work with a binomial logistic model.
• Make sure all riskratio_to_<change>() functions work with a binomial log-link model.
• Deprecate probs_to_odds(<data.frame>) and odds_to_probs(<data.frame>) - suggest something else instead.
• Tests
• Vignette?

[bwiernik]

Maybe "risk" instead of "probs"

[mattansb]

We already have odds_to_probs() - should that be renamed to odds_to_risk()?

[mattansb]

m1 <- glm(am ~ factor(cyl),
          family = binomial("logit"),
          data = mtcars)

m2 <- glm(am ~ factor(cyl),
          family = binomial("log"),
          data = mtcars)

effectsize::oddsratio_to_arr(m1)
#> Warning: 'p0' not provided:
#>   Computing effect size relative to the intercept (p0 = 0.727);
#>   Make sure your intercept is meaningful.
#> Parameter |   ARR |         95% CI
#> ----------------------------------
#> (p0)      |  0.73 |               
#> cyl [6]   | -0.30 | [-0.65,  0.11]
#> cyl [8]   | -0.58 | [-0.71, -0.22]
#> 
#> Uncertainty intervals (profile-likelihood) and p-values (two-tailed)
#>   computed using a Wald z-distribution approximation.
effectsize::riskratio_to_arr(m2)
#> Warning: 'p0' not provided:
#>   Computing effect size relative to the intercept (p0 = 0.727);
#>   Make sure your intercept is meaningful.
#> Parameter |   ARR |         95% CI
#> ----------------------------------
#> (p0)      |  0.73 |               
#> cyl [6]   | -0.30 | [-0.60,  0.23]
#> cyl [8]   | -0.58 | [-0.70, -0.30]
#> 
#> Uncertainty intervals (profile-likelihood) and p-values (two-tailed)
#>   computed using a Wald z-distribution approximation.

effectsize::oddsratio_to_riskratio(m1)
#> Warning: 'p0' not provided:
#>   Computing effect size relative to the intercept (p0 = 0.727);
#>   Make sure your intercept is meaningful.
#> Parameter | Risk ratio |       95% CI
#> -------------------------------------
#> (p0)      |       0.73 |             
#> cyl [6]   |       0.59 | [0.11, 1.16]
#> cyl [8]   |       0.20 | [0.02, 0.70]
#> 
#> Uncertainty intervals (profile-likelihood) and p-values (two-tailed)
#>   computed using a Wald z-distribution approximation.
parameters::model_parameters(m2, exp = TRUE)
#> Parameter   | Risk Ratio |   SE |       95% CI |     z |     p
#> --------------------------------------------------------------
#> (Intercept) |       0.73 | 0.13 | [0.44, 0.92] | -1.72 | 0.085
#> cyl [6]     |       0.59 | 0.28 | [0.17, 1.31] | -1.12 | 0.264
#> cyl [8]     |       0.20 | 0.13 | [0.03, 0.59] | -2.39 | 0.017
#> 
#> Uncertainty intervals (profile-likelihood) and p-values (two-tailed)
#>   computed using a Wald z-distribution approximation.

effectsize::riskratio_to_oddsratio(m2)
#> Warning: 'p0' not provided:
#>   Computing effect size relative to the intercept (p0 = 0.727);
#>   Make sure your intercept is meaningful.
#> Parameter | Odds ratio |       95% CI
#> -------------------------------------
#> (p0)      |       0.73 |             
#> cyl [6]   |       0.28 | [0.05, 7.54]
#> cyl [8]   |       0.06 | [0.01, 0.28]
#> 
#> Uncertainty intervals (profile-likelihood) and p-values (two-tailed)
#>   computed using a Wald z-distribution approximation.
parameters::model_parameters(m1, exp = TRUE)
#> Parameter   | Odds Ratio |   SE |        95% CI |     z |     p
#> ---------------------------------------------------------------
#> (Intercept) |       2.67 | 1.81 | [0.77, 12.17] |  1.45 | 0.147
#> cyl [6]     |       0.28 | 0.29 | [0.03,  1.99] | -1.24 | 0.214
#> cyl [8]     |       0.06 | 0.06 | [0.01,  0.39] | -2.72 | 0.007
#> 
#> Uncertainty intervals (profile-likelihood) and p-values (two-tailed)
#>   computed using a Wald z-distribution approximation.

^{Created on 2025-07-19 with reprex v2.1.1}

[bwiernik]

What did you mean by?

Deprecate probs_to_odds(<data.frame>) and odds_to_probs(<data.frame>) - suggest something else instead.

[mattansb]

There used to be a method for probs_to_odds() for converting selected columns in data frames between odds and probabilities, but that was using old {datawizard} selection syntax and was IMO a very specific use case that can easily be converted done with other tools. If a user now passes a data frame to probs_to_odds() they get this error:

Error: probs_to_odds(<data.frame>) is deprecated.
  You can use:
  datawizard::data_modify(data, .at = ..., .modify = probs_to_odds)
  Or
  dplyr::mutate(data, dplyr::across(..., probs_to_odds))
  Instead.