inter and intra event standard deviation and resiaduals using nlmefit function

Hi,

i am using nlmefit function to find regression coefficients for my mixed effect model.

for validating results i must check both inter and intra event standard deviation and resiaduals. how i can see these 4 final results?

is there any connection between stats.ires & stats.pres and inter & intra event resiaduals?

[the cyclist]

Only commenting because you asked me via email to take a look.

I've never heard the terms intra- and inter-event used in the context of a mixed effects model. I'm not sure I'll be able to help, but do you have any references to this terminology that you could cite?

[Mohammad Ebrahimi]

thank you so much for your help. I was really helpless, after about 2 months I couldn't find any answer to my question.

my data set is composed of earthquake information. and I want to find a mixed model for a special parameter using earthquake sources. so the grouping parameter will be each event and every event will have some records. so my residuals will be divided between events(inter-event)residual and with-in events(intra-event)residual.

the picture below is part of the paper I'm working on it.

[Tom Lane]

I have never heard the term inter- and intra-event residuals.

Your equation shows eta(i) and epsilon(i,j) as additive. This page:

Mixed-Effects Models - MATLAB & Simulink (mathworks.com)

does not show an equation with this form. It is something like

f(phi,x(i,j))+epsilon(i,j).

If the function had an additive form,

f(phi,x(i,j)) = g(phi0,x(i,j)) + eta(i)

where eta is one element of phi and phi0 is the rest of phi, and if you defined eta as random, then this would seem to match the form in the paper you provided. Then I would expect the estimated random effects eta to be the between-event residuals as you define them, and stats.ires would be the within-event residuals.

But I am not sure of this. Is there a simple published problem you can try? That is, can you find a worked-out example where these residuals are given, and see if you can match them with nlmefit?

[Mohammad Ebrahimi]

Thank you for your generous response.

The papers I have read used nlme package of R and some of them used Matlab but they haven't mentioned which function they have used in Matlab. Our function form is nonlinear so I think it must be nlmefit function.

According to the residual plots, inter-event residuals length must equal the number of groups and intra-event residuals length must equal the number of records. Also, it is mentioned that both residuals are normally distributed with zero means.

stats.pres is not normally distributed with zero means so it's not one of my residuals. stats.ires is normally distributed with zero means and its length is equal to the number of records not the number of groups. Also stats.ires is the difference between observed data and predicted data considering random effects. So I think stats.ires must equal to:

stats.ires = eta(i) + epsilon(i,j)

Actually, this random effect method is proposed by ABRAHAMSON and I have attached his paper in this comment. If you have any idea how to separate these two residuals from each other I will be glad if you let me know.

[Chao Feng]

Hi

Recently, I am having the same problem, may I ask you that did you solve the problem? thank you?