Possible problem with small alpha values (esr_backtest)

[EmanuelSommer]

Hi,

first of all thanks for the great package. Unfortunately I get an error that leaves me clueless for the code below with the test data that you can find here.

The code snippet:
esback::esr_backtest(test$r,test$q,test$e, alpha = 0.01, version = 1)
The error is:
Error in x %*% (bu - bl) : non-conformable arguments

Maybe you can help with this or at least clearify why there might be limitations.

Thanks a lot in advance!

[BayerSe]

Hi Emanuel,

thanks for reaching out and providing an example with your report.

The reason for this behavior is that the number of observations is too small for this type of model. If you check out our paper, you'll see that the smallest number of observations we consider in our simulations is 250 with a value of alpha of 5%.

With n=100 and alpha=1% you're out of luck here, sorry.

[EmanuelSommer]

Thanks for the quick reply!

Just to clarify the alpha refers to the probability level τ in the paper, right? So you actually used as the smallest specification n=250 and alpha=0.025 in the MC simulation on page 37 of your paper? Please correct me if I am wrong.

Just as a sidenote I also got the same error for sample sizes close to 300 and the alpha level of 1%. Input checks that trigger more informative error messages would be great in this context.

[BayerSe]

Sure, I'm glad someone is actually using the package ... ;-). And yes, I meant $\tau = 0.025$. Sorry for the confusion.

Thanks for the hint with the error messages. I think I should indeed add more checks on the input data. Thanks for bringing this up.