Years as smooths vs factors? #218
-
|
For the spatiotemporal models, is there ever a reason to include year as a smooth (rather than factor)? |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment
-
|
Yes, or as an IID random intercept effect or random walk or AR1 process. One reason that it is often included as a fixed effect factor is because we often want an independent estimate of the value of the response for each year (e.g., not constraining a population index for a stock assessment). We’ll have some examples with a smoother or random walk for years in the forecasting slides. |
Beta Was this translation helpful? Give feedback.
Yes, or as an IID random intercept effect or random walk or AR1 process. One reason that it is often included as a fixed effect factor is because we often want an independent estimate of the value of the response for each year (e.g., not constraining a population index for a stock assessment). We’ll have some examples with a smoother or random walk for years in the forecasting slides.
The smooth effects can also be seen as being akin to random effects – which will be shrunk to a common mean, and fixed effects won’t