CMDA 4654 Chapter Notes - Chapter Module 8 - Time Series: Simple Linear Regression, Independent And Identically Distributed Random Variables, Time Series

• School: Virginia Tech
• Department: Computational Modeling and Data Analytics
• Course: CMDA 4654
• Professor: R B Gramacy

Document Summary

Time series data are a collection of observations gathered over time. , yt could be annual gdp, quarterly production levels, weekly sales, daily temperature, etc. We might expect what happens at time t to be correlated with time t l where l is the lag. See if yt 1 is useful for predicting yt. To symmarize the time-varying dependence, compute lag-l correlations for l = 1, 2, 3, . This can be expressed in r by acf(dat). This provides a visual summary of the time- dependence in the data. The blue dashed lines denote statistical signi cance. r(l) = corr(yt, yt l) = ryt,yt l (1) Suppose y1 = 1, y2 = 1 + 2, . In a random walk model, the expected value of what will happen is always what happened most recently. Despite yt being a function of errors going to the beginning of the data, it can be written as depending only on yt 1.