Fish 550 Spring 2023 - 3 Team 1 - Lab 1

[giscus[bot]]

Fish 550 Spring 2023 - 3  Team 1 - Lab 1

https://atsa-es.github.io/fish550-2023/Lab-1/Final_Write_ups/Lab-1-team-1_final.html

[eeholmes]

Nice job comparing the FRI, actual and ARIMA forecasts. One thing I neglected to tell you is that the FRI (and AKFW) forecasts are one-year ahead, so when comparing ARIMA to FRI, one has do 1-year ahead ARIMA forecasts. So you'd keep adding 1-year to your training data so that your always have t-1 to do your ARIMA forecast.

It is interesting that the best model via AIC was wo a trend even for the data with a trend. Basically ARIMA(0,1,1) is saying that the best 1-step ahead forecast is the last observed data + some autocorrelation in the error.