Subscribe Bookmark RSS Feed
gail_massari

Community Manager

Joined:

Feb 27, 2013

In the Time Series GNP example, the R2 is 1. Doesn't that indicate overfitting?

In his JMP 12 Advanced Mastering JMP webcast on Time Series Analysis and Forecasting, Jian Cao buit a model where the R2 was 1.  He explains that for Times Series Analysis, where the data set is auto-correlated, R2 is not a good indicator for Goodness of Fit.  Rather, he suggests looking at AIC and SBC values, and examining how the model fits the historical data set and how well the model predicts.   He reminds us that JMP calculates many statistics in the background, some of which may not be relevant for the situation.  The user needs to assess the statistics based on some domain and content knowledge.

(view in My Videos)

 

Article Labels
Article Tags
Contributors