|
|
vendredi 29 mai 2015 à 18h31
|
The seasonal trend would be stacking together the numbers from one to about 320, once for each season, exactly the way the x-axis is labeled in the last plot.
The way I thought about this is that …
|
|
|
vendredi 29 mai 2015 à 16h40
|
`exog` can be any explanatory variables, the automatically included intercept and trend options are similar to it.
`exog` doesn't have to be a real explanatory variable like feed, but can also just…
|
|
|
vendredi 29 mai 2015 à 15h36
|
Ok, it wasn't clear to me how you combine different lactation "seasons"
The problem, as briefly discussed in the other thread, is that if you run ARIMA on the entire sample (4 seasons like in the …
|
|
|
vendredi 29 mai 2015 à 14h21
|
According to the documentation and the code the default for `typ` is `linear` not `level`.
about the trend:
I think either differencing or explicit modeling of it would work.
However, if you d…
|
|
|
vendredi 29 mai 2015 à 00h49
|
(Je ne parle pas beaucoup the francais. Je suis un developeur de statsmodels)
Si vous utilize un `diff` dans le model ARIMA(1, 1, 1), il a besoin de `typ=level` pour `predict` http://statsmodels.s…
|