Safe predictions from a generalized linear model

Safe predictions from a generalized linear model

# S3 method for glm
safe_predict(object, new_data, type = c("link", "conf_int",
  "response", "class", "prob"), ..., std_error = FALSE, level = 0.95,
  threshold = 0.5)

Arguments

object	A glm object returned from a call to stats::glm().
new_data	TODO
type	What kind of predictions to return. Which predictions are available depends on the family of object. "link" and "conf_int" are available for all families. "link" produces numeric predictions on the linear predictor scale. "conf_int" produces numeric predictions on the response scale and corresponding confidence bounds. "response" results in a numeric prediction on the response scale and is available for families: gaussian Gamma inverse.gaussian poisson quasipoisson quasi "class" results in hard class predictions and is only available for binomial and quasibinomial families "prob" results in class predictions for each class and is only available for binomial and quasibinomial families Default is "link".
...	Unused. safe_predict() checks that all arguments in ... are evaluated via the ellipsis package. The idea is to prevent silent errors when arguments are mispelled. This feature is experimental and feedback is welcome.
std_error	Logical indicating whether or not calculate standard errors for the fit at each point. Not available for all models, and can be computationally expensive to compute. The standard error is always the standard error for the mean, and never the standard error for predictions. Standard errors are returned in a column called .std_error. Defaults to FALSE.
level	A number strictly between 0 and 1 to use as the confidence level when calculating confidence and prediction intervals. Setting level = 0.90 correspondings to a 90 percent confidence interval. Ignored except when type = "conf_int" or type = "pred_int". Defaults to 0.95.
threshold	A number between 0 and 1 to use as a threshold for classification. When the class probability for the class corresponding to a positive event is greater than the threshold, the event will be classified as positive. Defaults to 0.5.

Details

For GLMs, standard errors can only be calculated when type = "link".