Is a PC prior with P(range < X) = 0.95 valid to indicate maximum likely range?

[Berin]

I have several repeated measures data sets where inclusion of a replicated spatial field improves WAIC by 15-70 points relative to both a model without a spatial field or one with a single spatial field, yet the range is strongly prior-sensitive and poorly defined (it increases dramatically with increasing prior values and approaches the maximum dimensions of my site).

The maximum plausible range is c. 50 based on my study site.

Is a PC prior with P(range < 50) = 0.95 a valid way of incorporating this belief?

This yields a posterior mean range of c. 25, which seems reasonable for my site. Using values of 10, 20, 30 and 40 instead of 50 also yields posterior mean ranges of c. 25 (suggesting the value is prior-insensitive)

I’m new to INLA and I’d be very grateful for confirmation that I’m on the right track, or else some advice on how to improve my model and obtain a meaningful estimate of the range .

Thanks,
Berin

[Elias T. Krainski]

It is better to use P(ranger < small.reference.range) = small.prob as suggested in the original paper where this prior was proposed:

https://www.tandfonline.com/doi/full/10.1080/01621459.2017.1415907

Please see the attached example.

Elias

--
You received this message because you are subscribed to the Google Groups "R-inla discussion group" group.
To unsubscribe from this group and stop receiving emails from it, send an email to r-inla-discussion...@googlegroups.com.
To view this discussion, visit https://groups.google.com/d/msgid/r-inla-discussion-group/79e140e6-e244-4094-9209-302e058ef671n%40googlegroups.com.