build an Markov probability model from multi-age prediction models

build an Markov probability model from multi-age prediction models

Usage

.disease_markov_model_for_stitch_model(
  markov,
  model,
  step,
  qbins,
  required_conditions,
  min_obs_for_estimate = 10
)

Arguments

markov

• the markov model computed for the next age (older). the states in this age will be mapped to the states in this markov layer

model

• prediction model (output of build_cross_validation_time_stitch_classification_models)

step

• time between prediction models

qbins

• quantile bin size of prediction score for which the markov model will define a state

required_conditions

• any filter to apply to the patients to filter out from model computation, for example limiting the time window

min_obs_for_estimate

• minimum of observations required to compute probability per sex/sbin. If minimum is not available, probability will be compuated using all data.

Value

a markov model, a list with the following members:

• model - matrix containing the probability for each quantile(score) bin to reach each of the target_classes provided in the oldest model.

• local.model - data.frame containing the probability for each quantile(score) bin to reach each of the quantile(score) bins of the next model by age.

• qbins - bins

• target - data frame containing the target bin for this age model (to be used as outcome for the younger age model)