Use Bayesian regression when you need to understand how multiple variables influence a continuous outcome while accounting for nested groups, such as students within schools or sales across regions. This method helps you separate individual patterns from group-level trends, providing a detailed view of your data's structure without complex math.
Understanding Group-Level Effects
In many datasets, rows aren't independent. If you are analyzing performance across different teams, the team itself influences the result. Bayesian regression uses random intercepts to account for these group differences, preventing them from biasing your primary findings.
By setting up a model that treats groups as variations around a central trend, you gain a clearer picture of whether an effect is consistent across the entire dataset or specific to certain groups.
Clearer Insights with Probabilistic Results
Instead of providing a single 'yes' or 'no' answer, this method returns a distribution of possible effects. We report the 'probability of direction,' which measures how likely an effect is to be positive or negative. This helps you understand the strength and reliability of your findings intuitively.
Additionally, we provide the 'Bayes R-squared,' which explains how well your model accounts for the variations in your data, including both the general trends and the specific group-level patterns.
Verifying Your Analysis
Because Bayesian analysis relies on simulations to reach conclusions, verifying that the model performed correctly is essential. We automatically calculate convergence diagnostics like 'r-hat' and 'ESS' for you.
These metrics confirm that the model found a stable solution. If these diagnostics indicate a problem, Lattice provides clear recommendations—such as increasing the number of samples—so you can trust the results generated by your analysis.
1 · Intent → method
An LLM picks bayesian_regression from a fixed catalog.
2 · Method → numbers
Deterministic Python engine runs the math. Same input → same output.
3 · Numbers → plain language
A second LLM translates the result into your domain’s vocabulary.
How is Bayesian regression different from standard linear regression?
Unlike standard regression, Bayesian regression provides a full range of possible values for your estimates, known as the posterior distribution. It also allows you to model group-level differences explicitly, which helps identify if trends vary across different categories.
What does the 95% HDI in the output mean?
The 95% Highest Density Interval (HDI) represents the range where the true effect is most likely to fall, based on your data. You can interpret this as a 95% probability that the actual value lies within these bounds.
Tool input schema
Schema for bayesian_regression not exported yet (run pnpm export:registry).