Data scientist: 90% of peers lack confidence in time series forecasting — a 5-concept primer

mdancho84 · x · 2026-09-05

Data science educator mdancho84 kicked off a mini-thread on time series forecasting, claiming 90% of data scientists lack confidence in the topic (it took him 3 years to master). The thread walks through 5 core concepts in plain language:

Time series analysis: interpreting trends, patterns and relationships in timestamped data.
Autocorrelation: correlation between a variable's current and past values.
Partial autocorrelation (PACF): strips out confounding effects of earlier lags.
Seasonal decomposition (STL): uses a LOESS smoother to split a series into trend, seasonal and residual components; handles arbitrary seasonality and enables outlier detection on residuals.
Forecasting: combining historical data, statistical models and expert judgment to reduce — not eliminate — future uncertainty.

It doubles as a funnel for his free workshop, but the explanations themselves are clear and useful for beginners.

Original post →

Data scientist: 90% of peers lack confidence in time series forecasting — a 5-concept primer

More from coding & agent