# Is it possible to define a parametric probabilistic model

**URL:** <https://persalys.discourse.group/t/is-it-possible-to-define-a-parametric-probabilistic-model/85>\
**Category:** Persalys usage\
**Created:** [April 11, 2022, 9:13am UTC](https://persalys.discourse.group/t/is-it-possible-to-define-a-parametric-probabilistic-model/85 "2022-04-11T09:13:33Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![Lolo](https://avatars.discourse-cdn.com/v4/letter/l/f475e1/32.png) [@Lolo](https://persalys.discourse.group/u/Lolo)\
**Post date:** [April 11, 2022, 9:13am UTC](https://persalys.discourse.group/t/is-it-possible-to-define-a-parametric-probabilistic-model/85/1 "2022-04-11T09:13:33Z")

</div>

Hi to the Persalys team,

I wonder if there is a way to define parametrics data in the probabilistic model window ?

Indeed, when I create a variable with the associated probabilistic law (normal for example), I can only define law characteristics with “numbers” (numericals values).

I would like to define this law characteristics by anothers parameters.

For example, define a normal variable “X” with parameters :

- mean “Xmean”

- standard deviation “Xstd”

This model definition make it possible to define the law charateristics in an input file for example. Or to define law characteristics with a more complexe formulation [Xmean = (a + b) \*\*c]

Besides, the parameters “Xmean” and “Xstd” could be define by anothers probabilistics law. For example :

- “Xmean” has a mean “Xmeanmean” and standard deviation “Xmeanstd”

- “Xstd” has a mean “Xstdmean” and standard deviation “Xstdstd”

(Data given with bootstrap approach for example)

Thanks you

Laurent

---

<div class="post-metadata">

**Author:** ![dumas](https://yyz2.discourse-cdn.com/free1/user_avatar/persalys.discourse.group/dumas/32/5_2.png) [@dumas](https://persalys.discourse.group/u/dumas)\
**Post date:** [April 12, 2022, 4:37pm UTC](https://persalys.discourse.group/t/is-it-possible-to-define-a-parametric-probabilistic-model/85/2 "2022-04-12T16:37:58Z")

</div>

Hello,

Unfortunately it is not possible to do that, as it was in Phimecasoft I imagined you think. I think the only possibility is to do the math before or try directly in the Python model to create the final X random realization.

Regards,  
Antoine

---

<div class="post-metadata">

**Author:** ![Lolo](https://avatars.discourse-cdn.com/v4/letter/l/f475e1/32.png) [@Lolo](https://persalys.discourse.group/u/Lolo)\
**Post date:** [April 13, 2022, 6:13am UTC](https://persalys.discourse.group/t/is-it-possible-to-define-a-parametric-probabilistic-model/85/3 "2022-04-13T06:13:11Z")

</div>

Hi Antoine,

Ok, thanks you.  
Yes, this is a way to define variables in PhimecaSoft… Variables I would like to transpose in Persalys.

Indeed, a solution I found is to generate a random realization with a fortran code just before to create my input program (coupling model) with a fortran code too. This sequence is driven by a shell code in an upper level. Persalys generate the others variables.

I tried to make a python model but I found it difficult to write the python code and to make the link with my coupling program… for now !

Laurent