Trend forecasting Study Guide

📖 Core Concepts Trend Analysis – Collecting historical data to spot a consistent direction of change (pattern). Purpose – Use past behavior to predict future events or outcomes. Project‑Management Role – A mathematical technique that forecasts cost‑ and schedule‑performance based on earlier results; serves as a quality‑control tool. Statistical Goal – Extract the underlying pattern in a time‑series that is hidden by random noise. Linear vs. Non‑Linear Trends – Linear: Can be modeled with ordinary regression. Non‑Linear: Require non‑parametric tests (e.g., Mann‑Kendall) or smoothing methods. Extrapolation – Extending a known trend beyond the observed data to estimate future values. 📌 Must Remember Trend analysis = historical data → pattern → future prediction. In project management it monitors cost variance and schedule variance over time. Linear trend → use regression analysis. Non‑linear trend → use Mann‑Kendall test or smoothing techniques. The Mann‑Kendall test is a version of the Kendall rank correlation coefficient. Extrapolation is the step that turns a identified trend into a forecast. 🔄 Key Processes Collect Data – Gather time‑ordered observations (costs, schedule dates, market ratios, etc.). Visual Inspection – Plot the series to see if a pattern looks linear or curved. Choose Method If points lie roughly on a straight line → linear regression. If the pattern is curvilinear or noisy → non‑parametric test (Mann‑Kendall) or smoothing. Fit Model / Test – Run regression or Mann‑Kendall test; obtain slope or trend‑significance. Validate – Check residuals for randomness; ensure model assumptions hold. Extrapolate – Extend the fitted line or smoothed curve to the desired future horizon. 🔍 Key Comparisons Linear Trend vs. Non‑Linear Trend Linear → straight‑line fit, simple slope, assumes constant rate of change. Non‑Linear → curved pattern, needs Mann‑Kendall or smoothing, handles varying rates. Trend Analysis vs. Forecasting Trend analysis = identify the direction; Forecasting = apply the identified trend to predict specific future values. ⚠️ Common Misunderstandings “All trends are linear.” – The outline explicitly notes non‑linear trends and appropriate tests. “Trend analysis = forecasting.” – Trend analysis is the diagnostic step; extrapolation (a separate step) turns it into a forecast. “Project‑management trends only track cost.” – Both cost and schedule variances are monitored. 🧠 Mental Models / Intuition “Time‑Series Telescope” – Imagine looking through a telescope: the past points line up to show the direction you’ll see next. If the stars (data) form a straight line, you can point straight ahead (linear). If they curve, you adjust the angle gradually (non‑linear). “Noise vs. Signal” – Think of a radio: the underlying song is the trend; static is noise. Smoothing is like turning up the volume of the song while turning down the static. 🚩 Exceptions & Edge Cases Insufficient Data – Too few points make any trend unreliable; avoid extrapolation. Structural Breaks – Sudden policy, market, or technology shifts can flip a trend; regression slope may become meaningless. Seasonality – Regular periodic swings (e.g., weather) must be removed before trend testing, otherwise the Mann‑Kendall test can misinterpret cycles as trends. 📍 When to Use Which Cost/Schedule Monitoring (Project Management) → Use simple linear regression if performance changes steadily; switch to Mann‑Kendall if you see irregular jumps. Financial Ratio Trend → Compare ratios year‑over‑year; if the plot is jagged, apply smoothing (moving average) before interpreting. Technology or Weather Forecasts → Often involve non‑linear dynamics → start with smoothing or Mann‑Kendall, then extrapolate. 👀 Patterns to Recognize Straight‑Line Plot → Likely a linear trend → regression suffices. Consistent Up/Down Curvature → Hint of a non‑linear trend → consider Mann‑Kendall or smoothing. Repeated Up‑Down Fluctuations → Seasonal pattern → remove before trend analysis. Sudden Shift in Slope → Possible structural break → re‑evaluate the model period. 🗂️ Exam Traps Choosing Linear Regression for Curved Data – The answer will look clean but will be penalized for ignoring non‑linearity. Confusing “Trend” with “Forecast” – Selecting “trend analysis predicts exact numbers” is wrong; prediction requires extrapolation. Assuming Mann‑Kendall works on any data – It is only appropriate when you need a non‑parametric test for monotonic trends; using it on clearly linear data wastes effort. Over‑extrapolating Beyond Reasonable Horizon – Answers that extend a trend far beyond the data range are often distractors. --- Use this guide to quickly recall what trend analysis is, when to apply each method, and how to avoid the most common pitfalls on the exam.