Weird std.dev. values

Forums

RSS (Opens New Window)

Audric Vigier, modified 5 Years ago.

Weird std.dev. values

Youngling Posts: 6 Join Date: 9/6/16 Recent Posts

Hi all,

I'm working on a spatially-explicit SS3 model for european Northern hake, and trying to describe migrations in the model. Thus, after a run, I want to know what are the movement estimates, and their std dev. I read the standard deviation for each movement paremeter, to be sure these std dev are not too high. (in the SS_output object, data frame $parameters, Parm_StDev column).

For one movement parameter, bounded between -20 and 12, std dev is estimated at around 3000, which doesn't seem possible. I read the ParmTrace file to see if the parameter went out of bounds while the optimisation process, which could explain such a high std dev value, but it always stayed inside the bounds. Is there something I'm missing about std dev values calculation, or could it be a glitch?

Richard Methot, modified 8 Years ago.

RE: Weird std.dev. values

Youngling Posts: 222 Join Date: 11/24/14 Recent Posts

Hi Audric,

The bounds on a parameter will not directly influence the stddev of that parameter's estimate. A very large stddev could occur if the parameter has no prior and it has negligible influence on the model.

Sometimes when a model configuration has a Hessian that is barely invertable, all the parameters will get the same stddev, but that does not seem to be the problem here.

You might try fixing that parameter at a few different values and see if you get the same result regardless of the value.

Rick