STATS 489 Lecture Notes - Lecture 5: Time Series, Stock Market
Document Summary
When you look at results of a study, ask: why did they find this correlation, how did they find this correlation. It is very difficult to find a strong relationship over a long period of time: most of the graphs only showed a strong correlation over about 10 years. When you have a time series plot, the changes from year to year are usually not independent: super bowl and stock market are probably not independent, more likely to have high correlation. Pearson"s correlation: measures linear relationship, ranges from -1 to 1, if it"s equal to -1 or 1, then it"s a straight line. Spearman"s correlation: measures monotonic relationships, first convert data into ranks (from 1 to 3) Then calculate correlation: if your relationship is linear, it"s better to use pearson"s correlation because it"s more specific. Because the proportions of the outside factors are unbalanced, aggregating the data often results in a reversal of the correlation.