statistical-analysis

To calculate descriptive statistics and detect anomalies, the Skill computes mean, median, standard deviation, IQR, and percentiles, then applies Z-score and IQR-based methods to identify point anomalies and time-series deviations for investigation.

For hypothesis testing in A/B experiments, it suggests appropriate tests like t-tests or proportions z-tests, reports p-values, effect sizes, and confidence intervals, and advises on sample size limits and multiple-comparison corrections.

Trend analysis for time series monitoring works by using moving averages and YoY/MoM comparisons to establish baselines, check for seasonality, and generate simple short-term projections for your business metrics.

You need to escalate statistical analysis to data science when hitting complex scenarios like confounding, Simpson's paradox, or survivorship bias, as the Skill provides practical guidance on identifying these pitfalls and escalation criteria.

Yes, you can use this for nonparametric alternatives in hypothesis testing, as it suggests appropriate tests beyond standard t-tests and ANOVA, while reporting effect sizes and clear guidance on uncertainty.

To avoid overinterpreting business metrics, the Skill applies statistical methods that emphasize uncertainty and practical significance, providing confidence intervals, percentiles, and warnings about common pitfalls like survivorship bias.