Topic: Graph crashes when inserting certain trendlines

Ivan,

I'm using the latest beta version, but I know I have seen this happen in older versions as well.

Download the file and insert a trendline for the data. For the model, paste this in:

(($v-1)*$k*x+0.008208886^(1-$v))^(1/(1-$v))

Don't just click okay yet. This will give a division by zero error. I realize that this is normal. I assume this is because the program contains (1-$v) in a denominator, and the initial values of the parameters are set to 1 so we can never start fitting a curve. To avoid this, paste the formula and click Add Model. Give the variables reasonable starting values. I suggest k=1.8 and v=3. These values correspond to the curve already graphed, and this curve is already a reasonably good fit, so it should be a good place to start.

Click okay and try the fit. I get "Abnormal  Program Termination" and graph quits.

Also, using the built-in "exponential association" curve fit I sometimes get "Error evaluating power function" and sometimes "Access violation at address 007C3F1C in module 'Graph.exe'...." or even "External exception EEFACE". which is the worst because graph will not exit thereafter (you just get the same error again.)

If you just open the file and try to fit the exponential association trendline to the data, you get one error the first time, then another if you try again, and the third time it just exits. Very odd. Any idea what's going on here?

Post's attachments

Attachment icon Desorption Curve.grf 1.31 kb, 454 downloads since 2011-04-21 

Re: Graph crashes when inserting certain trendlines

Unfortunately I am not able to reproduce the problem. I have tried with the newest beta version (build 496) under both Windows XP and Windows 7 and I cannot reproduce the crash, neither with the custom model nor with the exponential fit.

Graph gives me this trendline: f(x)=((2.9051-1)*1.37*x+0.0082^(1-2.9051))^(1/(1-2.9051))
Unfortunately that makes it a little difficult to fix. I would like to hear if anyone else have seen this problem?