User Tools

Site Tools


fitting_a_model

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revision Previous revision
fitting_a_model [2016/11/28 15:42]
vincenzo
fitting_a_model [2016/11/28 15:43] (current)
vincenzo
Line 136: Line 136:
 </​code>​ </​code>​
  
-Let's try to add some other covariates - interaction terms between time bins and day. To evaluate the model we could check the R squared and RMSE. The R squared is not a very good measure since one can always artificially inflate it (R squared is how much of the variance is explained) by adding extra covariates. We can rely on the **RMSE** which makes a good general purpose error metric for numerical predictions. In this case, we can simply ​call summary(model)$ sigma for printing ​the metric. In the second model, the error is in fact smaller (0.9463 vs. 0.9433)+Let's try to add some other covariates - interaction terms between time bins and day. To evaluate the model we could check the R squared and RMSE. The R squared is not a very good measure since one can always artificially inflate it (R squared is how much of the variance is explained) by adding extra covariates. We can rely on the **RMSE** which makes a good general purpose error metric for numerical predictions. In this case, we can simply ​print the metric. In the second model, the error is in fact smaller (0.9463 vs. 0.9433)
  
 <code python> <code python>
fitting_a_model.txt ยท Last modified: 2016/11/28 15:43 by vincenzo