Define a missing-predictor model — impute

impute_model() wraps the model for a predictor used inside mi(). A bare formula in impute, such as impute = list(x = x ~ z), is still treated as a Gaussian model for a numeric missing predictor. Use impute_model() when the missing predictor needs an explicit non-Gaussian predictor family. The first non-Gaussian fitted routes are fixed-effect Bernoulli/logit models for one binary predictor, fixed-effect cumulative-logit models for one ordered categorical predictor, fixed-effect baseline-category softmax models for one unordered categorical predictor, fixed-effect beta models for one strict proportion predictor in (0, 1), fixed-effect zero-one beta models for one boundary proportion predictor in [0, 1], fixed-effect beta-binomial models for one denominator-aware success/trial proportion predictor, fixed-effect Poisson, negative-binomial, or zero-truncated negative-binomial models for one count predictor, fixed-effect lognormal or Gamma models for one positive continuous predictor, and fixed-effect Tweedie models for one non-negative semi-continuous predictor with exact zeros. Most current non-Gaussian predictor families are fitted inside a Gaussian response location model; Poisson, binomial, negative-binomial, and beta responses are currently supported for one binary missing predictor.

Usage

impute_model(formula, family = stats::gaussian(), trials = NULL)

Arguments

formula: Two-sided predictor-model formula. For most families, the left-hand side must be the same variable used inside mi(). For family = beta_binomial(), the left-hand side is the success-count column, while the mi() variable is the success proportion used in the response model.
family: Predictor-model family. gaussian() keeps the existing continuous predictor route. binomial(link = "logit") fits the binary missing-predictor route. cumulative_logit() fits the ordered categorical missing-predictor route. categorical() fits the unordered categorical missing-predictor route. beta() fits the strict beta/proportion missing-predictor route. zero_one_beta() fits the boundary-proportion missing-predictor route. beta_binomial() fits a denominator-aware success/trial proportion route and requires trials. poisson(link = "log"), nbinom2(), and truncated_nbinom2() fit count missing-predictor routes. lognormal() and Gamma(link = "log") fit positive continuous missing-predictor routes. tweedie() fits a semi-continuous non-negative missing-predictor route with exact zeros.
trials: Optional trial-count column for family = beta_binomial(). The formula left-hand side is the success count and trials is the known denominator for each row.

Value

A drm_impute_model object for the impute argument of drmTMB().

Examples

impute_model(x ~ z)
#> $formula
#> x ~ z
#> <environment: 0x556a9b1d15d8>
#> 
#> $family
#> 
#> Family: gaussian 
#> Link function: identity 
#> 
#> 
#> $family_type
#> [1] "gaussian"
#> 
#> $trials
#> NULL
#> 
#> attr(,"class")
#> [1] "drm_impute_model"
impute_model(treatment ~ z, family = binomial())
#> $formula
#> treatment ~ z
#> <environment: 0x556a9b1d15d8>
#> 
#> $family
#> 
#> Family: binomial 
#> Link function: logit 
#> 
#> 
#> $family_type
#> [1] "bernoulli"
#> 
#> $trials
#> NULL
#> 
#> attr(,"class")
#> [1] "drm_impute_model"
impute_model(score ~ z, family = cumulative_logit())
#> $formula
#> score ~ z
#> <environment: 0x556a9b1d15d8>
#> 
#> $family
#> $name
#> [1] "cumulative_logit"
#> 
#> $family
#> [1] "cumulative_logit"
#> 
#> $n_response
#> [1] 1
#> 
#> $dpars
#> [1] "mu"
#> 
#> $links
#>         mu 
#> "identity" 
#> 
#> attr(,"class")
#> [1] "drm_family"
#> 
#> $family_type
#> [1] "ordinal"
#> 
#> $trials
#> NULL
#> 
#> attr(,"class")
#> [1] "drm_impute_model"
impute_model(habitat ~ z, family = categorical())
#> $formula
#> habitat ~ z
#> <environment: 0x556a9b1d15d8>
#> 
#> $family
#> $name
#> [1] "categorical"
#> 
#> $family
#> [1] "categorical"
#> 
#> $link
#> [1] "baseline_softmax"
#> 
#> attr(,"class")
#> [1] "drm_impute_family"
#> 
#> $family_type
#> [1] "categorical"
#> 
#> $trials
#> NULL
#> 
#> attr(,"class")
#> [1] "drm_impute_model"
impute_model(cover ~ z, family = beta())
#> $formula
#> cover ~ z
#> <environment: 0x556a9b1d15d8>
#> 
#> $family
#> $name
#> [1] "beta"
#> 
#> $family
#> [1] "beta"
#> 
#> $n_response
#> [1] 1
#> 
#> $dpars
#> [1] "mu"    "sigma"
#> 
#> $links
#>      mu   sigma 
#> "logit"   "log" 
#> 
#> attr(,"class")
#> [1] "drm_family"
#> 
#> $family_type
#> [1] "beta"
#> 
#> $trials
#> NULL
#> 
#> attr(,"class")
#> [1] "drm_impute_model"
impute_model(cover ~ z, family = zero_one_beta())
#> $formula
#> cover ~ z
#> <environment: 0x556a9b1d15d8>
#> 
#> $family
#> $name
#> [1] "zero_one_beta"
#> 
#> $family
#> [1] "zero_one_beta"
#> 
#> $n_response
#> [1] 1
#> 
#> $dpars
#> [1] "mu"    "sigma" "zoi"   "coi"  
#> 
#> $links
#>      mu   sigma     zoi     coi 
#> "logit"   "log" "logit" "logit" 
#> 
#> attr(,"class")
#> [1] "drm_family"
#> 
#> $family_type
#> [1] "zero_one_beta"
#> 
#> $trials
#> NULL
#> 
#> attr(,"class")
#> [1] "drm_impute_model"
impute_model(success ~ z, family = beta_binomial(), trials = trials)
#> $formula
#> success ~ z
#> <environment: 0x556a9b1d15d8>
#> 
#> $family
#> $name
#> [1] "beta_binomial"
#> 
#> $family
#> [1] "beta_binomial"
#> 
#> $n_response
#> [1] 1
#> 
#> $dpars
#> [1] "mu"    "sigma"
#> 
#> $links
#>      mu   sigma 
#> "logit"   "log" 
#> 
#> attr(,"class")
#> [1] "drm_family"
#> 
#> $family_type
#> [1] "beta_binomial"
#> 
#> $trials
#> trials
#> 
#> attr(,"class")
#> [1] "drm_impute_model"
impute_model(abundance ~ z, family = poisson())
#> $formula
#> abundance ~ z
#> <environment: 0x556a9b1d15d8>
#> 
#> $family
#> 
#> Family: poisson 
#> Link function: log 
#> 
#> 
#> $family_type
#> [1] "poisson"
#> 
#> $trials
#> NULL
#> 
#> attr(,"class")
#> [1] "drm_impute_model"
impute_model(abundance ~ z, family = nbinom2())
#> $formula
#> abundance ~ z
#> <environment: 0x556a9b1d15d8>
#> 
#> $family
#> $name
#> [1] "nbinom2"
#> 
#> $family
#> [1] "nbinom2"
#> 
#> $n_response
#> [1] 1
#> 
#> $dpars
#> [1] "mu"    "sigma"
#> 
#> $links
#>    mu sigma 
#> "log" "log" 
#> 
#> attr(,"class")
#> [1] "drm_family"
#> 
#> $family_type
#> [1] "nbinom2"
#> 
#> $trials
#> NULL
#> 
#> attr(,"class")
#> [1] "drm_impute_model"
impute_model(abundance ~ z, family = truncated_nbinom2())
#> $formula
#> abundance ~ z
#> <environment: 0x556a9b1d15d8>
#> 
#> $family
#> $name
#> [1] "truncated_nbinom2"
#> 
#> $family
#> [1] "truncated_nbinom2"
#> 
#> $n_response
#> [1] 1
#> 
#> $dpars
#> [1] "mu"    "sigma"
#> 
#> $links
#>    mu sigma 
#> "log" "log" 
#> 
#> attr(,"class")
#> [1] "drm_family"
#> 
#> $family_type
#> [1] "truncated_nbinom2"
#> 
#> $trials
#> NULL
#> 
#> attr(,"class")
#> [1] "drm_impute_model"
impute_model(biomass ~ z, family = lognormal())
#> $formula
#> biomass ~ z
#> <environment: 0x556a9b1d15d8>
#> 
#> $family
#> $name
#> [1] "lognormal"
#> 
#> $family
#> [1] "lognormal"
#> 
#> $n_response
#> [1] 1
#> 
#> $dpars
#> [1] "mu"    "sigma"
#> 
#> $links
#>         mu      sigma 
#> "identity"      "log" 
#> 
#> attr(,"class")
#> [1] "drm_family"
#> 
#> $family_type
#> [1] "lognormal"
#> 
#> $trials
#> NULL
#> 
#> attr(,"class")
#> [1] "drm_impute_model"
impute_model(biomass ~ z, family = Gamma(link = "log"))
#> $formula
#> biomass ~ z
#> <environment: 0x556a9b1d15d8>
#> 
#> $family
#> 
#> Family: Gamma 
#> Link function: log 
#> 
#> 
#> $family_type
#> [1] "gamma"
#> 
#> $trials
#> NULL
#> 
#> attr(,"class")
#> [1] "drm_impute_model"
impute_model(biomass ~ z, family = tweedie())
#> $formula
#> biomass ~ z
#> <environment: 0x556a9b1d15d8>
#> 
#> $family
#> $name
#> [1] "tweedie"
#> 
#> $family
#> [1] "tweedie"
#> 
#> $n_response
#> [1] 1
#> 
#> $dpars
#> [1] "mu"    "sigma" "nu"   
#> 
#> $links
#>        mu     sigma        nu 
#>     "log"     "log" "logit12" 
#> 
#> attr(,"class")
#> [1] "drm_family"
#> 
#> $family_type
#> [1] "tweedie"
#> 
#> $trials
#> NULL
#> 
#> attr(,"class")
#> [1] "drm_impute_model"