Curve fitting: Taking standard deviation into account

[benodin]

thanks...

especially for the weighting (ups )

(this particular gaussian fit is usless for me because I need mine. Even it looks very good :))

[altenbach]

The weight is 1/variance, so you need to square the sdtdev. i.e. weight=(1/sd)² . (see also). Now the origin and nonlinear fit are very similar, see attached.

I don't understand why you have a big sequence structure. There is no problem with execution order.

---

Also most of your code is extremely convoluted. As an example, let's have a look at how you generate the origin data (see picture)

A simple autoindexing FOR loop is all that's needed (middle). You could even use an expression node, pasting the formula "as is". (I don't like that express VI at all, because you never know what it's doing unless you open the panel.)

[altenbach]

Here's a "slightly" cleaned-up version. Modify as needed.

LabVIEW is parallel. Don't chop up the code with sequence structures. I would also use complex for xy graphs, simpler data structures. 😄

[benodin]

Unbelievable, thanks a lot.

I appreciate that.

And many thanks for the weighting issue as well!

(The sequence structure was only to show what I'm doing in a clear way)

[benodin]

Ha, but I found an other question.

Now I get values for a,b,c,d. How do I get the standard deviation of them?

THANKS

[altenbach]

---

@benodin wrote:

Now I get values for a,b,c,d. How do I get the standard deviation of them?

---

You can get it from the covariance matrix. Start with this thread.

[benodin]

as far as I understand:

take each value of the diagonal -> take the square root and multiply with the RMSE-> std of the coefficients

at least the results look very reasonable