POL242Y5 Lecture Notes - Lecture 18: Logistic Regression, Statistical Significance, Moving Average
Document Summary
Autocorrelation: when you have the same units over time, say observe same country cannot attempt to interpret it individually: period variation. Certain time period that it repeats: example 8 am tim"s more busy then at 2 in the afternoon, random variation, stuff bounces around. Average of the past n values: 3 day average, drop old 3rd day and new day, replace with new, always lags events, but it always lags if there is a sudden change, it will take 3 days. Exponential moving average more sensitive to recent data before data realize drastic change: weight newer data more, older data less, first day . 1, second day . 3, third day . 6, more sensitivity, but les smoothing. If you want to periodize data together but need to know what"s the cycle period, need to know that in advance. Remove time period: compare the same periods, de-mean the data.