How to contribute new methods

Alex Hayes

2018-12-26

what needs to happen in each safe_predict method

check for suggested packages if you rely on those

reference implementations

see safe_predict.lm() for example

document if you successfully deal with:

• missing data

new_data validation is model specific

• are necessary predictors present
• what input types are okay

input edge cases

- spline trap

tests

the zoo of formulas to use to test the formula interface

## function signature

## input validation works

# - error on missing `new_data`
# - error on bad `type`
# - error on bad `std_error`
# - error on bad `level` - include 0 or 1
# - warning on mispelled argument (via ellipsis)

## input edge cases

# - missing data
# - spline trap
# - single observation
# - repeated observations

## checks on returned predictions

# - correct class given type
# - in a tibble
# - correct number of rows
# - level and interval attributes for type = conf_int and type = pred_int
# - correct column names given type