## 14.10 Missing data

One nice features of this approach is that it is robust to a fair bit of missing data. Here I delete a third of the data. I do this randomly throughout the dataset. The data look pretty hopeless. No signal to be seen.

Fit as usual:

fit <- MARSS(dat2.miss, model = mod.list1, silent = TRUE)

But though we can’t see the signal in the data, it is there.

Averaging our sensors doesn’t work since there are so many missing values and we will have missing values in our average.

Another type of missing data are strings of missing data. Here I create a data set with random strings of missing values. Again the data look really hopeless and definitely cannot average across the data since we’d be averaging across different data sets.

We can fit as usual and see that it is possible to recover the signal.

fit <- MARSS(dat2.miss, model = mod.list1, silent = TRUE)