Question about NNETAR from Section 12.4, Hyndman, 3rd edition

Machine Learning and Modeling
1

Hello:

Please help with with the syntax of NNETAR. The example from 12.4 works, but when I try to modify it, I get an error:

fit_seasonal <- sunspots |> model(NNETAR(sqrt(value), formula = AR(p = NULL, P = NULL, period = 11)))
Error in list2(...) : object 'value' not found

I tried to specify p and P explicitly below. It doesn't fail, but the fitted model doesn't correspond to the specification:

fit_seasonal <- sunspots |> model(NNETAR(sqrt(value), p = 1, P = 1, period = 11))
fit_seasonal |> report()

Series: value
Model: NNAR(9,5)

Here I try to specify the period and find p and P automatically:

fit_seasonal <- sunspots |> model(NNETAR(value ~ AR(p = NULL, P = NULL, period = 11)))
Warning message:
1 error encountered for NNETAR(value ~ AR(p = NULL, P = NULL, period = 11))
[1] missing value where TRUE/FALSE needed

Referred here by Forecasting: Principles and Practice, by Rob J Hyndman and George Athanasopoulos

2 
library(fpp3)
sunspots <- sunspot.year |> as_tsibble()
fit <- sunspots |>
  model(NNETAR(sqrt(value) ~ AR(P = 1, period = 11)))

However, I wouldn't use this model for sunspots. They are not periodic, and forcing a period of 11 doesn't really make sense. They are cyclic, not seasonal.