Analyze a single round of observed data¶
This guide covers the use case where you already have one round of A/B/n data from your own platform and do not need pytyche’s sequential allocation loop — you want a posterior fit and a decision, once.
The canonical one-shot path is pt.analyze(observed), which runs fit
auto-selection via pt.fit then immediately calls .analyze() on the
resulting posterior. For callers who need access to the intermediate
posterior object, the explicit two-step alternative is equivalent:
posterior = pt.fit(observed)
result = posterior.analyze()
Both paths produce the same AnalysisResult; the one-liner is convenience
sugar for the common case where you have no reason to inspect the raw
posterior draws before producing a decision.
For a full walkthrough of the posterior object and what you can do with it
before calling .analyze(), see the
working with the posterior
tutorial. For the types returned by .analyze(), see
result objects.