STA 4753 Lecture 3: Time-series Analysis. Intro to Forecasting
Document Summary
It"s useful to examine observations over time in order to predict (forecast) future values. Examine one variable over time (a time series) Two-step process: build a model that fits the data observed (historical, use the model to predict (forecast) future values. Time-series: a sequence of ordered values of a variable in equally spaced time intervals. Helps in understanding any underlying forces that produced the observed data. Are used to fit a model and forecast values that allow us to monitor and provide feedback. Series with trends- observations fluctuate regularly through time. Series with seasonality- observations are high, then drop off a repeating pattern through time periods. Noise: unpredictable component that gives a time-series graph its zig-zag appearance. Cross sectional data: measurements on multiple units that are recorded in a simple period of time. Panel data: cross sectional measurements repeated over time periods. Data observed over time doesn"t always compare to each other because: