Aggregation Use Case: Standalone Aggregation

This use case demonstrates running the Aggregation module directly against a pre-computed statistics file, independent of the MET Stat or Statistics components.

Prerequisites

Before running the example, make sure VCasT is installed. Follow the installation steps in Quick Start Guide.

Run the Example

Clone the test repository:

git clone https://github.com/NOAA-GSL/VCasT-tests
cd VCasT-tests/examples/Significance/ss-0002

Run VCasT with the provided YAML file:
```
vcast config_agg.yaml
```
This reads ../input_files/modelb-T2M.data, groups rows by fcst_lead, and averages the rmse and bias columns within each group.

YAML Configuration Explained

Sample aggregation configuration

# ====================================================
# Aggregation Configuration File
# ====================================================
# This YAML configuration is used to run the aggregation mode.
# It specifies the input file containing all the processed statistics,
# the columns by which the data should be grouped/aggregated, and
# the output path for saving the aggregated results.

# ------------------------
# Input Data File
# ------------------------
input_file: "/path/to/inputfile.data"       # Path to the processed input data file containing all statistics

# ------------------------
# Grouping Settings
# ------------------------
group_by: ["model", "fcst_lead"]            # List of columns used for aggregation.
                                             # The order determines how the data is grouped.

ci: true                                    # Add Student's t-based confidence intervals

stats:
  - 'rmse'
  - 'bias'

# ------------------------
# Output Data File
# ------------------------
output_agg_file: "/path/to/output_agg.data"   # Path and filename to save the aggregated data

Output

The output (agg.data) contains one row per unique fcst_lead value, with a count column and one column per requested statistic (rmse, bias) holding the group mean. Set ci: true in the configuration to additionally emit {stat}_bcl / {stat}_bcu confidence-interval bound columns for each statistic.