I have recently come across a few suggestions to enhancements to the Data Analysis Window that seem reasonable and tie in to older requests with new insights.
First, in this post http://forum.geogebra.org/viewtopic.php?f=13&t=27635 it is suggested that the old issue of logarithmic axes might be resolved by restricting this much wanted option to the Data Analysis Window, thereby eliminating the need to make all internal objects "logarithmic aware". With this restriction, It is probably doable. It is only points, functions, axes and grids that need to know how to draw themselves in a logarithmic fashion.
Second, in the Norwegian Facebook group, a discussion recently unearthed the need to have general regression with the Fit command incorporated in this window so that regressions of any type (i.e. y = ax, y = c e^kx + b...) could be fit within this window. That would mean perhaps a new check box for "more models" and an input box for a model function or a list of functions. There is no need for the definitions of these to be made in the Data Analysis Window, thay can be made in "normal GeoGebra mode".
In addition to these, I wish to re-iterate an old wish of mine that ties in nicely with the last one:
In this post http://forum.geogebra.org/v... I wish for standard errors of the coefficients of regression to be calculated. It was said it was planned for 5.0, I hope it is still on the to-do-list. The Data Analysis Window would pe a useful place to display these errors, though commands to calculate them would perhaps be a good complement.
As I have stated earlier, this last feature would make error analysis in physics lab reports so much more accessible to students, to say nothing of those studying error propagation who could then dynamically see how the errors vary when individual points (representing measurements) are moved in relation to others.
The Data Analysis Window is already a great tool. E.g. the ability to dynamically select which data points are active, makes it simple to use bootstrapping techniques to approximate errors. But no tool is so great that it cannot be made awesome :wink:
I like this idea 
If these changes was made I would use Geogebra as the only tool in my teaching.
An observation:
For a while now, Desmos, has been able to do regressions nicely. Part of what makes it so easy is that there is no need to first define a model function and then issue the regression command. I believe the original request in this post addresses this issue. The possibility to enter any function and assume that undefined variables are parameters to fit the function to the data would make advanced regression really easy. Another nifty thing Desmos does is that they define a list for the residuals, meaning you can fit another, second order function to to them. Perhaps this functionality is something to consider for a future upgrade of the Data Analysis tool.
I can see that this is getting close now. It is in the beta-version!
It would be nice, if the resulting curve was always a graph of a function.
It would also be nice, if the user could change the axis labels from within the Data Analysis Window.
Comments have been locked on this page!