Riccardo, thanks for replying. 

Let me provide more information , as I reckon I did a bad job the first time ;)

Thanks for the welcoming!

The tricky bit of analysis will handle angles, distances and deltas of those variables, since everything needs to be georeferenced, every observation takes a coordinate pair (X,Y or N,E) but also Z (X,Y,Z). 

The data will take the form of a panel model, with a benchmark value for the base variables (the average of a few dozen observations ) and the recurring measurements shall be compared against the benchmark, and once enough data is collected, we'll try to understand various relationships amongst variables and their observed values.

My emphasis on the (XYZ) is that though rain, temp, exogenous impacts, etc play a role in the *seasonal* observations therefore recurring, the coordinates are observed down to a high precision (mm) and changes in them may flag changes in the observed values which are not related to any recurring cycle but a deterioration in the condition of a specific location, and many tests and hypothesis will be checked for each observation. There geo variable should be so interrelated that I expected 

y= a + b*sin(x) is a generalization for an accumulation of a geographical position, such as Y = Yi + d*sine(alpha) for example. When I run an OLS the results seem to explain very little of what I expected to be a straightforward and linear expression, specially because it's a panel and I am comparing the same data which is expected to change very little between observations/measurements.

So first of all I need to script for checking 1mm change in coordinates (above or below each variable X, Y, Z) then check for season impacts etc.

I am attaching an example file with test measurements. Important detail (in order to make sense of data) is that I am creating a BENCHMARK value for XYZ (average value) and creating variables DeltaX/Y/Z which are the ones I will compare against 0.001m (1mm) change by doing ex. Delta_X(i) = X(i) - mean(X) etc. I also should say that the dataset used a less precsa equipment so it is very hard to check for 1mm changes, as per OLS , the results will tell.

My idea is to use tests to check if the changes breach a critical value and if they relate to seasonal factors (no big deal) or they are indeed red flags.

Any ideas how you'd run this?

Cheers from Brazil!

Em sex., 8 de dez. de 2023 às 11:51, Riccardo (Jack) Lucchetti <p002264@staff.univpm.it> escreveu:

On 08/12/2023 14:52, Meridiana GeoTopo wrote:
> Hello Gretl Enthusiasts!
>
> I hope this message finds you well. I am reaching out to this
> knowledgeable community as I embark on a unique project that involves
> utilizing Gretl for the estimation of environmental and engineering
> parameters. While I've already dived into Gretl for this purpose, the
> project's diversity from traditional econometrics poses a few challenges.

Welcome to the community!

> Specifically, I am looking to run regression models where I can create
> custom formulas and expressions. To give you a glimpse, my requirements
> involve working with a Cartesian coordinate system, something like y = a
> + d(sine(b)), as I heavily rely on spatial data in my study and also
> structural health.

Some clarification is needed here. If (as I believe form the context) a
and b are parameters and d is an observable series, then all you need is
OLS plus the application of the delta method in order to recover the
estimate of b and its standard error. Otherwise, if b is an observable
series, then sin(b) is an observable series too, and a distinction must
be made according to what the "d()" expression means. If it means "a
parameter called d multiplied by something", then OLS suffices, again.
Otherwise, if it means "the function d() applied to something", where
this function continuous and differentiable, then what you need is NLS.

All these methods are provided by gretl, so feel free to ask for guidance.

> I could also use TLS and GLS models - can I run them in gretl and if so
> where can I find examples of implementations?

The TLS model as such is not implemented in gretl natively, since it's
very seldom used (if ever) in econometrics. However, we do have a
function for performing the SVD and it shouldn't be difficult to write a
script to perfom TLS estimation. As for GLS, it really depends on what
you need (GLS is a very generic term that could be applied to many
different things). If what you need is the special case known in
econometrics as "weighted least squares", then yes, we do have a "wls"
command.

> Later we will attach soil, air, temperature, rain etc to the model but
> these will follow a linear model approach easily.
>
> As a newcomer to Gretl, I am seeking honest assistance, guidance, and
> insights from the experienced members of this group. If anyone could
> spare a moment to share your expertise, provide script help, or guide me
> on using the GUI to achieve these objectives, I would greatly appreciate it.
>
> Your support will not only contribute significantly to the success of my
> project but will also help me enhance my understanding of Gretl in a
> practical context. I am eager to learn and absorb insights from the
> Gretl community.
>
> Thank you in advance for your time and assistance. I look forward to the
> valuable input from this esteemed group.

Feel free to ask!

-------------------------------------------------------
   Riccardo (Jack) Lucchetti
   Dipartimento di Scienze Economiche e Sociali (DiSES)

   Università Politecnica delle Marche
   (formerly known as Università di Ancona)

   r.lucchetti@univpm.it
   http://www2.econ.univpm.it/servizi/hpp/lucchetti
-------------------------------------------------------
_______________________________________________
Gretl-users mailing list -- gretl-users@gretlml.univpm.it
To unsubscribe send an email to gretl-users-leave@gretlml.univpm.it
Website: https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/