Rework score()

DESCRIPTION

@@ -53,6 +53,7 @@ Imports:
     ggplot2 (>= 3.4.0),
     lifecycle,
     methods,
+    Metrics,
     rlang,
     scoringRules,
     stats

NAMESPACE

@@ -2,6 +2,16 @@
 
 S3method(plot,scoringutils_available_forecasts)
 S3method(print,scoringutils_check)
+S3method(score,default)
+S3method(score,scoringutils_binary)
+S3method(score,scoringutils_point)
+S3method(score,scoringutils_quantile)
+S3method(score,scoringutils_sample)
+S3method(validate,default)
+S3method(validate,scoringutils_binary)
+S3method(validate,scoringutils_point)
+S3method(validate,scoringutils_quantile)
+S3method(validate,scoringutils_sample)
 export(abs_error)
 export(add_coverage)
 export(ae_median_quantile)
@@ -13,11 +23,10 @@ export(bias_quantile)
 export(bias_range)
 export(bias_sample)
 export(brier_score)
-export(check_forecasts)
 export(correlation)
 export(crps_sample)
 export(dss_sample)
-export(find_duplicates)
+export(get_duplicate_forecasts)
 export(interval_score)
 export(log_shift)
 export(logs_binary)
@@ -26,6 +35,7 @@ export(mad_sample)
 export(make_NA)
 export(make_na)
 export(merge_pred_and_obs)
+export(new_scoringutils)
 export(pairwise_comparison)
 export(pit)
 export(pit_sample)
@@ -50,13 +60,25 @@ export(summarise_scores)
 export(summarize_scores)
 export(theme_scoringutils)
 export(transform_forecasts)
+export(validate)
+export(validate_general)
+importFrom(Metrics,ae)
+importFrom(Metrics,ape)
+importFrom(Metrics,se)
 importFrom(checkmate,assert)
 importFrom(checkmate,assert_data_frame)
+importFrom(checkmate,assert_data_table)
 importFrom(checkmate,assert_factor)
+importFrom(checkmate,assert_list)
 importFrom(checkmate,assert_numeric)
 importFrom(checkmate,check_atomic_vector)
+importFrom(checkmate,check_data_frame)
+importFrom(checkmate,check_function)
 importFrom(checkmate,check_matrix)
 importFrom(checkmate,check_numeric)
+importFrom(checkmate,test_factor)
+importFrom(checkmate,test_list)
+importFrom(checkmate,test_numeric)
 importFrom(data.table,"%like%")
 importFrom(data.table,':=')
 importFrom(data.table,.I)
@@ -71,6 +93,7 @@ importFrom(data.table,melt)
 importFrom(data.table,nafill)
 importFrom(data.table,rbindlist)
 importFrom(data.table,setDT)
+importFrom(data.table,setattr)
 importFrom(data.table,setnames)
 importFrom(ggdist,geom_lineribbon)
 importFrom(ggplot2,.data)
@@ -113,7 +136,6 @@ importFrom(ggplot2,xlab)
 importFrom(ggplot2,ylab)
 importFrom(lifecycle,deprecated)
 importFrom(methods,hasArg)
-importFrom(methods,is)
 importFrom(rlang,enexprs)
 importFrom(rlang,warn)
 importFrom(scoringRules,crps_sample)

NEWS.md

@@ -1,32 +1,35 @@
 # scoringutils 1.3
 
-This major update introduces a lot of breaking changes and addresses comments made by review from the Journal of Statistical Software (see preprint of the manuscript [here](https://arxiv.org/abs/2205.07090)).
+This major update and addresses a variety of comments made by reviewers from the Journal of Statistical Software (see preprint of the manuscript [here](https://arxiv.org/abs/2205.07090)).
+
+The update introduces breaking changes. If you want to keep using the older version, you can download it using `remotes::install_github("epiforecasts/scoringutils@v1.2")`. 
 
 ## Package updates
-- in `score()`, required columns "true_value" and "prediction" were replaced by required columns "observed" and "predicted". Scoring functions now also use the function arguements "observed" and "predicted" everywhere consistently. 
-- scoring functions received a consistent interface and input checks:
-  - metrics for binary forecasts: 
+- In `score()`, required columns "true_value" and "prediction" were renamed and replaced by required columns "observed" and "predicted". Scoring functions now also use the function arguments "observed" and "predicted" everywhere consistently. 
+- Scoring functions received a consistent interface and input checks:
+  - metrics for binary forecasts:
     - `observed`: factor with exactly 2 levels
     - `predicted`: numeric, vector with probabilities
-  - metrics for point forecasts: 
+  - metrics for point forecasts:
     - `observed`: numeric vector
     - `predicted`: numeric vector
-  - metrics for sample-based forecasts: 
+  - metrics for sample-based forecasts:
     - `observed`: numeric, either a scalar or a vector
     - `predicted`: numeric, a vector (if `observed` is a scalar) or a matrix (if `observed` is a vector)
-  - metrics for quantile-based forecasts: 
+  - metrics for quantile-based forecasts:
     - `observed`: numeric, either a scalar or a vector
     - `predicted`: numeric, a vector (if `observed` is a scalar) or a matrix (if `observed` is a vector)
-    - `quantile`: numeric, a vector with quantile-levels. Can alternatively be a matrix of the same shape as `predicted`. 
-- changes to `avail_forecasts()` and `plot_avail_forecasts()`:
-  - the function `avail_forecasts()` was renamed to `available_forecasts()` for consistency with `available_metrics()`. The old function, `avail_forecasts()` is still available as an alias, but will be removed in the future. 
-  - For clarity, the output column in `avail_forecasts()` was renamed from "Number forecasts" to "count". 
-  - `available_forecasts()` now also displays combinations where there are 0 forecasts, instead of silently dropping corresponding rows. 
-  - `plot_avail_forecasts()` has been deprecated in favour of an S3 method for `plot()`. An alias is still available, but will be removed in the future. 
-- remove hard-coded rounding value for `correlation()`. Previously, the function always rounded correlations to two digits. Instead, a new argument, `digits` was introduced and the default set to 0, meaning that no rounding takes place. 
-- the deprecated `..density..` was replaced with `after_stat(density)` in ggplot calls.
-- files ending in ".Rda" were renamed to ".rds" where appropriate when used together with `saveRDS()` or readRDS()`. 
-- add documentation for the return value of `summarise_scores()`. 
+    - `quantile`: numeric, a vector with quantile-levels. Can alternatively be a matrix of the same shape as `predicted`.
+- `check_forecasts()` was replaced by a new function `validate()`. `validate()` validates the input and in that sense fulfills the purpose of `check_forecasts()`. It has different methods: `validate.default()` assigns the input a class based on their forecast type. Other methods validate the input specifically for the various forecast types.
+- The function `find_duplicates()` was renamed to `get_duplicate_forecasts()`
+- Changes to `avail_forecasts()` and `plot_avail_forecasts()`:
+  - The function `avail_forecasts()` was renamed to `available_forecasts()` for consistency with `available_metrics()`. The old function, `avail_forecasts()` is still available as an alias, but will be removed in the future.
+  - For clarity, the output column in `avail_forecasts()` was renamed from "Number forecasts" to "count".
+  - `available_forecasts()` now also displays combinations where there are 0 forecasts, instead of silently dropping corresponding rows.
+  - `plot_avail_forecasts()` has been deprecated in favour of an S3 method for `plot()`. An alias is still available, but will be removed in the future.
+- The deprecated `..density..` was replaced with `after_stat(density)` in ggplot calls.
+- Files ending in ".Rda" were renamed to ".rds" where appropriate when used together with `saveRDS()` or `readRDS()`.
+- added documentation for the return value of `summarise_scores()`. 
 
 # scoringutils 1.2.1
 

R/available_forecasts.R

@@ -37,10 +37,9 @@ available_forecasts <- function(data,
                                 by = NULL,
                                 collapse = c("quantile", "sample")) {
 
-  check_data <- check_forecasts(data)
-
-  data <- check_data$cleaned_data
-  forecast_unit <- check_data$forecast_unit
+  data <- validate(data)
+  forecast_unit <- attr(data, "forecast_unit")
+  data <- remove_na_observed_predicted(data)
 
   if (is.null(by)) {
     by <- forecast_unit
@@ -90,4 +89,3 @@ avail_forecasts <- function(data,
   )
   available_forecasts(data, by, collapse)
 }
-