TSF Study DeskUnit 1 + Unit 2

Last pass

Revision That Actually Sticks

Use this after reading the units. It is designed for the night before: recall first, then repair weak spots.

90-Minute Plan

  1. 0-15 min: read the cheat sheet definitions and formulas aloud.
  2. 15-35 min: write the components and decomposition answers from memory.
  3. 35-60 min: solve one odd-window and one even-window moving average problem.
  4. 60-75 min: rehearse exponential smoothing types and model-selection logic.
  5. 75-90 min: attempt five practice questions without looking at answers.

Oral Rehearsal

Forecasting

Forecasting estimates future values using available information. It supports planning, decision making, risk reduction, and resource allocation.

Time series

A time series is ordered observations of the same variable over time. Order and spacing matter because observations are dependent.

Components

Trend is long-run movement, seasonality is fixed-period repetition, cyclical movement has no fixed period, and irregular variation is random noise.

Smoothing

Moving average uses equal weights inside a window. Exponential smoothing gives more weight to recent observations.

Formula Dump

Yt = Tt + St + It
Yt = Tt x St x It
y_hat_(t+1) = yt
MA = sum(window values) / n
SES = Ft + alpha(yt - Ft)
F_(t+h) = Lt + hBt
MAE = average(|et|)
MAPE = average(|et / yt|) x 100

Numerical Flow

Problem typeFirst stepFinal check
Odd moving averageTake rolling sums and divide by n.Values align to middle year.
Even moving averageCompute raw moving averages.Average adjacent raw MAs to center them.
Least-squares lineCode x values around 0.Substitute the target year's coded x.
SESChoose or identify initial forecast.Use same alpha consistently down the table.
HoltInitialize level and trend.Forecast uses level plus h times trend.

Self-Test

Say yes only if you can do it without notes