Update Curve Fitter data

I'm using MatLab R2023a which comes with an "improved" layout for the Curve Fitting Toolbox

Let me describe my workflow, and my problem. I have a script that generates data into variables x and y. In a previous version of the Curve Fitting tool, I could select the dropdown and re-select x and y in order to get my new data points.

Now, with the updated version, this takes several more steps. I have to first open the select data dialog box, which is not super quick to load, then I have to select some variable that isn't x or y, because now if I select the same variable that I had before, it doesn't update the data set, and nothing happens. Then I have to reselect x and y in the select data fields, and finally close the dialog box.

Alternatively, I could lauch a new session using curveFitter(x,y), however, then I have to re-configure all of my fit parameters.

Does anyone have a suggestion? I have to do a lot of fits, and this is seriously interrupting my workflow. I'm considering downloading an old version of MatLab and using that instead because of this issue.

[dpb]

Can you not write a script/function to do the work without the GUI interface getting in the way? I've never really seen the need for it for other than maybe a first-time purely exploratory peek at a new dataset, but certainly not for anything repetitive or production-like...

[James Lambert]

@dpb My problem is that for many of my curve fits, I end up needed to exclude a fairly significant number of points. I think you're right though, I need to find a way to fit my data that doesn't involve manually excluding points. I'm very much not an expert at doing fits, so maybe this was the kick I needed to re-analyze and try to figure out how to fit my data better.

[dpb]

If the data aren't so sensitive as to be unable to be made available, I'd suggest attaching a couple representative sets with the kind of model you think appropriate. Having some context surrounding data is generally of benefit in help folks have relevant ideas on appropriate models. May entice some regulars here to chime in; often there's the challenge can't resist... :)

[dpb]

I'd also suggest submitting the complaint/observation about changes in functionality as a Bug/Quality of Implementation issue to get it in the official complaints channel. TMW staff may/may not see any given post in Answers; that makes it an official complaint.

[James Lambert]

Thanks you @dpb, I did figure out a way to fit my set of data without repeating steps in the curve fitting toolbox, but I think I will also submit an observation as well, because it can still be annoying at times. One thing that also made it more bareable was turning off auto-fit, that way it doesn't take the time to process each fit every time you select a data source

[dpb]

Agree -- enhancement requests/GUI foopahs and the like are always worth submitting...I even submit suggestions for changes to help files when find overlooked details that are significant or the like.

Answer 1

[Steven Lord]

If you're using the Curve Fitter app you could fit your first set of data interactively then generate code from the fit and use that code for subsequent programmatic fits. See the fourth example on that documentation page.

Some of the data preprocessing Live Editor Tasks or the Data Cleaner app could help you prepare your data for use with Curve Fitter. See this documentation page for a list of available tasks; if you're excluding outliers from your data I'd take a look at the Clean Outlier Data task. Like most if not all Live Editor Tasks the Clean Outlier Data task generates code that you could incorporate into the code generated from Curve Fitter.

[James Lambert]

Thank you, I'll look into these options