[R] Time series (trend over time) for irregular sampling dates and multiple sites
@purd|e@@ @end|ng |rom gm@||@com
Wed May 1 06:06:49 CEST 2019
This is possibly off topic now...
However, given that it involves mgcv, I think that it's relevant to R.
> to test if there is a change over the years on the amount of debris in
these locations and more specifically a change after the implementation of
a mitigation strategy
> My debris items per effort (Ieffort) are fishing and shipping related
items that can be due to an intentional discharge or an accidental
discharge. It is very common to find a great amount of these items together
in the beach (from where we collected these data (beach clean-ups),
possibly having origin from the same ship. I was thinking that this can be
a problem but still don't know how to overcome or if it makes sense to
include in the model.
I could be wrong on this.
If your goal is simply to determine whether the MARPOL term in significant
or not (or how strong the effect is), I don't think the above issue is
However, you could do a separate spatial analysis, which could be very
> This does not apply along the different years.
Are you sure (there's no long term effect)?
Note that you could combine Year and nMonth into one variable, say t.
However, if I understand your variables correctly, this would be correlated
So, if you try to fit a model with both Year and DaysIa, then Year is less
likely to be significant, and you probably don't need both.
Note that another approach, is to regard month as a categorical variable.
Also, note that it may be worthwhile testing for interactions, between
MARPOL and Location or Site.
If you want to be fancy, you could test for interactions between MARPOL and
your time variables.
It's possible that there are higher order interactions, however, these sort
of models are difficult for most people to interpret, so are probably a bad
[[alternative HTML version deleted]]
More information about the R-help