CHAOSS models

The repometrics package allows the construction of “models” derived by the CHAOSS community (“Community Health Analytics in Open Source Software”). These models collate various numbers of individual “metrics”:

to provide deeper context and answer more complex questions about a community’s health.

The metrics used in the models are described in the separate CHAOSS metrics vignette, and are detailed in the data.frame returned by the rm_chaoss_metrics_list() function.

In contrast to the metrics, each of which is encoded as a separate function within this package, the models are not hard-coded, rather the metrics used in each are defined within a single JSON file stored at extdata/chaoss-models/chaoss-models.json in the inst sub-directory of the package. This allows the model construction to be easily modified without needing to modify the underlying code of this package.

The file can be accessed by loading the package:

library (repometrics)

and then locating the file with:

f <- system.file (
    fs::path ("extdata", "chaoss-models", "chaoss-models.json"),
    package = "repometrics"
)

This file defines all CHAOSS metrics used in this package, and lists the metrics used to construct each model.

The models used here

• Developer responsiveness
• Project engagement
• Project awareness
• Community activity
• OSS compliance and security
• OSS project viability: community
• OSS project viability: starter
• OSS project viability: governance
• OSS project viability: strategy
• Collaboration development index
• Community service and support
• Starter project health
• Community welcomingness

Each model collates individual metrics defined in the chaoss-models.json file, with final model values summing individual values of all component metrics, as described in the following sub-section.

Combining metrics within models

Different CHAOSS metrics quantify different properties, and are measured on different scales. To combine metrics into a single model, differences in scale must be reconciled. The chaoss-models.json file defines each metrics according to one of five types of measurement:

• binary, translated to 0 for false and 1 for true;
• count, for integer numbers greater than or equal to 0;
• ratio, for floating-point numbers between 0 and 1;
• days, for integer measurements in days; and
• real, for single, real values greater than zero.

Metrics measured as counts or days are log-transformed before being aggregated to final models. Each metric also indicates where higher or lower values are desirable. Aggregate model scores are designed so that higher values are better, so any metrics for which lower values are better are first negated before being aggregated into overall model scores.