Forecasting-6-4-MREG-relative-importance.Rmd

## Relative importance metrics {#MREGRELPO}

<!--
if(file.exists("Fish-Forecast.Rmd")) file.remove("Fish-Forecast.Rmd")
bookdown::preview_chapter("Forecasting-6-2-Covariates-MSEG.Rmd")
-->

Another approach to linear regression with multiple collinear regressors is to compute relative importance metrics[^relimp]. The **relaimpo** package will compute the relative importance metrics and provides plotting.

[^relimp]: Groemping, U. (2006) Relative Importance for Linear Regression in R: The Package relaimpo Journal of Statistical Software 17, Issue 1. Downloadable at http://www.jstatsoft.org/v17/i01

This gives a somewhat different picture with year, Trachurus and the effort metrics most important while the environmental variables have low importance.

```{r relimp}
reli <- relaimpo::calc.relimp(anchovy~.,data=df)
plot(reli)
```

The pattern remains the same without Year as a response variable.

```{r relimp.noYear}
reli <- relaimpo::calc.relimp(anchovy~.-Year,data=df)
plot(reli)
```