Using E-Surge for individual covariates

[Guillaume Souchay]

Hello,

I'd like to put some individual covariates (BCI on the first age class) in a multisite model, it's not possible with M-Surge, so I'm going to use E-Surge for a multievent model where there is no error in the attribution of states.
I'd like to know if there was some explanation about how to input some covariate and the format of the file, is it the same as population covariates?

Kind regards,

Guillaume

[Guillaume Souchay]

I have the answer thanks to Rémi Choquet: there are some documents on the "help" top menu of E-Surge and one is about individual covariates.

Guillaume

[Guillaume Souchay]

Hi,
after some problems with my data, finally I can use E-Surge with individual covariates. But I have a problem: E-Surge found no available data:
Subscribed assignment dimension mismatch

Error in ==> saisie at 120

Error in ==> menumuse>LoaddataM_Callback at 457

Error in ==> gui_mainfcn at 75

Error in ==> menumuse at 42

Error while evaluating uimenu Callback

Is there some format for the covariate? number of terms after coma?
Is it possible to let some NA data?
I'm using some residuals of type II regression (weight x tarsus).

2 lines for exemple:
1000000000000400000000400 1 0.00540472 ;
1000000000000004000000000 1 -0.0373708 ;

Kind regards,

Guillaume

[CHOQUET]

Hello,

E-SURGE don't accept the Na which were put as a value for the
missing co-variates. Use instead 0.

Missing co-variates are not presently allowed in E-SURGE.
This is a current challenge in the modelling and we are thinking
about this problem.

Rémi

[Guillaume Souchay]

Hi Rémi,
in the model I'm using, I use the word "xind" to set an covariate effect.
The covariate is in my data set, in the last column.
For the model, I entered:
f.a(2,3:6,7)+f(1).a(1).xind+others
but in the IVFV windows, I don"t have the "others" parameter.
Is it possible to set the covariate only on the survival from 1 at age 1 and to fix the other survival at age 1 to 1?

Thanks,

Guillaume

[CHOQUET]

Normally yes !!!!

[Guillaume Souchay]

Ok,
so is the sentence : f.a(2,3:6,7)+f(1).a(1).xind+others correct?
I have the individual covariate in the data set and I noticed this to the logiciel and it asked for covariate in the data.
But I can't find a parameter 211111 (from 2 to 1 at occasion 1 for age 1 group 1 for step 1) in order to fix it to 1 in the IVFV window.
I only have a parameter "fix ind cov" and parameter 111111 or 221111.

Thanks,

Guillaume

[CHOQUET]

> so is the sentence : f.a(2,3:6,7)+f(1).a(1).xind+others correct?

No, this sentence is incorrect, use instead

f.a(2,3:6,7)+f(1).a(1).[i+xind]+f(2).a(1)

So that, there is one intercept link to parameter f(1).a(1).

[Guillaume Souchay]

Hi,

Thanks for the lastest answer, I could have some precious results.
a second hypothesis is that BCI covariate could have some effects on the survival on older age than age 1, so I tried to compute a modem with the following sentence for survival:

f(1).a(1)+f(2).a(1)+f.a(3:6,7)+f.a(2).[i+xind]

I use this sentence rather than
f(1).a(1)+f.a(3:6,7)+f.a(2).[i+xind]+others
to avoid some problems between "xind" and "others".

Using this sentences, E-Surge had an error and had to close:

??? Error using ==> vertcat
CAT arguments dimensions are not consistent
Error in ==> modme.contraint at 130
Error in ==> menumuse> Gemaco_Callback at 600
Error in ==> gui_mainfcn at 96
Error in ==> menumuse at 42

Is the used sentence correct or not? (And so is the origin of the error)

Thanks a lot!

[CHOQUET]

Hello,

sorry for this very late reply.
This error was due to a bug in GEMACO. The consequence is that
the redundancy is not fully well treated when there is an individual covariate as in our sentence:

f(1).a(1)+f(2).a(1)+f.a(3:6,7)+f.a(2).[i+xind]

to avoid the bug, use the following sentence instead:

f(1).a(1)+f(2).a(1)+f.a(3:6,7)+f(1,2).a(2).[i+xind]

The bug will be corrected in the next version of E-SURGE.

Sorry again,

Rémi Choquet