If you attempt to use multiplicative seasonality on a series with negatives (e.g., monthly temperature), the error message that comes back from seasonplot is highly uniformative. There's no logical checks in decomposition to make sure that the series is suitable for multiplicative, so the logarithm gives NaN's, and stl dies.
Suggest: some logical checks at the start of decomposition() to give informative error messages.
If I have time, I'll submit a pull request in a couple of weeks that fixes this issue.
If you attempt to use multiplicative seasonality on a series with negatives (e.g., monthly temperature), the error message that comes back from seasonplot is highly uniformative. There's no logical checks in decomposition to make sure that the series is suitable for multiplicative, so the logarithm gives NaN's, and stl dies.
Suggest: some logical checks at the start of decomposition() to give informative error messages.
If I have time, I'll submit a pull request in a couple of weeks that fixes this issue.