Changelog

insurancerating (development version)

Model data extraction (`extract_model_data()` / `model_data()`)

Added extract_model_data() to retrieve cleaned model data from glm, refitsmooth, and refitrestricted objects.
model_data() is deprecated as of 0.8.0 and now emits a lifecycle warning; it remains as a wrapper.

update_smoothing()

Introduces a dedicated helper to update existing smoothing specifications without refitting the full model from scratch.
Enables faster, more transparent iteration when fine-tuning smoothing curves.

Rating table output (`rating_table()` / `rating_factors*()`)

Renamed rating_factors() to rating_table() as the primary user-facing API.
rating_factors() and rating_factors2() are deprecated as of 0.8.0 and now emit lifecycle warnings; they remain as wrappers.

Histogram functions (`outlier_histogram()` / `histbin()`)

Added outlier_histogram() as the new name for histograms with outlier bins.
histbin() is deprecated as of 0.8.0 and now warns; it remains as a wrapper.
The x argument must now be provided as a string (standard evaluation).
No functional changes to binning or outlier handling.

Breaking changes

The function fit_gam() has been deprecated and replaced by riskfactor_gam().
- fit_gam() used non-standard evaluation (NSE), allowing unquoted column names.
- riskfactor_gam() now uses standard evaluation (SE), requiring column names as character strings.
  Example migration:
```
# old (NSE, deprecated)
fit_gam(df, nclaims = nclaims, x = age_policyholder, exposure = exposure)

# new (SE)
riskfactor_gam(df, nclaims = "nclaims", x = "age_policyholder", exposure = "exposure")
```
- The NSE wrapper fit_gam() is still available but will show a deprecation warning and will be removed in a future release.
The function univariate() has been deprecated and replaced by univariate_summary().
- univariate() used non-standard evaluation (NSE), allowing unquoted column names.
- univariate_summary() now uses standard evaluation (SE), requiring column names as character strings.
  Example migration:
```
# old (NSE, deprecated)
univariate(df, x = area, severity = amount, nclaims = nclaims, exposure = exposure)

# new (SE)
univariate_summary(df, x = "area", severity = "amount",
                   nclaims = "nclaims", exposure = "exposure")
```
- The NSE wrapper univariate() is still available but will show a deprecation warning and will be removed in a future release.

Minor changes

Improved documentation for riskfactor_gam(), including clearer examples and migration guidance from fit_gam().

insurancerating 0.7.5

CRAN release: 2024-10-09

rating_factors() now always returns correct output when column with exposure in data is not named exposure
intercept_only in update_glm() is added to apply the manual changes and refit the intercept, ensuring that the changes have no impact on the other variables.
smoothing in smooth_coef() is added to choose smoothing specification
The README has been revised

insurancerating 0.7.4

CRAN release: 2024-05-20

bootstrap_rmse() now uses after_stat(density) instead of the deprecated dot-dot notation
custom_theme in autoplot.univariate() is added to customize the theme

insurancerating 0.7.3

CRAN release: 2024-05-09

autoplot.univariate() now generates a plot even when there are missing values in the rows
rating_factors() now always returns the correct coefficients when used on a ‘refitsmooth’ or ‘refitrestricted’ class of GLM.

insurancerating 0.7.2

CRAN release: 2022-12-20

update_glm() now always returns the correct interval in case the function is used in combination with smooth_coef()

insurancerating 0.7.1

CRAN release: 2022-09-06

rotate_angle in autoplot.univariate() is added to rotate x-labels
univariate() now accepts external vectors for x; vec_ext() must be used

insurancerating 0.7.0

CRAN release: 2022-07-08

smooth_coef() now gives correct results for intervals with scientific notation
reduce() now returns no errors anymore for columns with dates in POSIXt format

insurancerating 0.6.9

CRAN release: 2021-12-11

refit_glm() is renamed to update_glm()
construct_model_points() and model_data() are added to create model points

insurancerating 0.6.8

CRAN release: 2021-11-10

show_total in autoplot.univariate() is added to add line for total of groups in case by is used in univariate(); total_color can be used to change the color of the line, and total_name is added to change the name of the legend for the line
rating_factors() now accepts GLMs with an intercept only
fit_truncated_dist() is added to fit the original distribution (gamma, lognormal) from truncated severity data
join_to_nearest() now returns NA in case NA is used as input

insurancerating 0.6.7

CRAN release: 2021-07-28

smooth_coef() now returns an error message when intervals are not obtained by cut()
get_data() is added to return the data used in refit_glm()

insurancerating 0.6.6

CRAN release: 2021-05-19

summary.reduce() now gives correct aggregation for periods “months” and “quarters”
rows_per_date() is added to determine active portfolio for a certain date

insurancerating 0.6.5

CRAN release: 2021-03-22

smooth_coef() and restrict_coef() are added for model refinement
histbin() now uses darkblue as default fill color

insurancerating 0.6.4

CRAN release: 2021-01-12

In summary.reduce(), name can be used to change the name of the new column in the output.
Dataset MTPL now contains extra columns for power, bm, and zip.
Some functions in insight are renamed, therefore insight::format_table() is replaced with insight::export_table().

insurancerating 0.6.3

CRAN release: 2020-10-28

fit_gam() for pure premium is now using average premium for each x calculated as sum(pure_premium * exposure) / sum(exposure) instead of sum(pure_premium) / sum(exposure) (#2).
histbin() is added to create histograms with outliers
reduce now returns a data.frame as output

insurancerating 0.6.2

CRAN release: 2020-06-08

check_normality() is now depreciated; use check_residuals() instead to detect overall deviations from the expected distribution
rating_factors() now shows significance stars for p-values
period_to_months() arithmetic operations with dates are rewritten; much faster
univariate() now has argument by to determine summary statistics for different subgroups

insurancerating 0.6.1

CRAN release: 2020-04-29

univariate_all() and autoplot.univ_all() are now depreciated; use univariate() and autoplot.univariate() instead
check_overdispersion(), check_normality(), model_performance(), bootstrap_rmse(), and add_prediction() are added to test model quality and return performance metrics
reduce() is added to reduce an insurance portfolio by merging redundant date ranges

insurancerating 0.6.0

CRAN release: 2020-04-10

label_width in autoplot() is added to wrap long labels in multiple lines
sort_manual in autoplot() is added to sort risk factors into an own ordering
autoplot() now works without manually loading package ggplot2 and patchwork first
rating_factors() now returns an object of class riskfactor
autoplot.riskfactor() is added to create the corresponding plots to the output given by rating_factors()

insurancerating 0.5.2

CRAN release: 2020-03-30

autoplot.univ_all() now gives correct labels on the x-axis when ncol > 1.

insurancerating 0.5.1

CRAN release: 2020-03-29

A package website is added using pkgdown.
construct_tariff_classes() and fit_gam() now only returns tariff classes and fitted gam respectively; other items are stored as attributes.
univariate_frequency(), univariate_average_severity(), univariate_risk_premium(), univariate_loss_ratio(), univariate_average_premium(), univariate_exposure(), and univariate_all() are added to perform an univariate analysis on an insurance portfolio.
autoplot() creates the corresponding plots to the summary statistics calculated by univariate_*.

insurancerating 0.5.0

CRAN release: 2020-03-12

construct_tariff_classes() is now split in fit_gam() and construct_tariff_classes().
A vignette is added on how to use the package.

insurancerating 0.4.3

CRAN release: 2019-11-01

period_to_months() is added to split rows with a time period longer than one month to multiple rows with a time period of exactly one month each.

insurancerating 0.4.2

CRAN release: 2019-05-31

In construct_tariff_classes(), model now also accepts ‘severity’ as specification.