How to solve this error

Question

Yosuke Nakatani 2023-10-31

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2040761-how-to-solve-this-error

回答： Arthur Goldsipe 2023-10-31

To estimate parameter values for my model, fit program was used in model analyzer (Simbiology).

Parameter was successfully estimated when I choosed uniform weght ([]) in the program. However, if I changed weight option (e.g. 1/[DV1 DV2]), program didn't work and got an error message ("All values in the weights matrix must be positive, real, and finite.").

I'd like to know how to solve this problem (which is the problem model or data?)

2 个评论
显示无隐藏无

Walter Roberson 2023-10-31

If DV1 or DV2 are 0 then 1 divided by them would be infinite which would be a problem.

Yosuke Nakatani 2023-10-31

Hi Walter,

Thank you. I input zero for DV1 at time 0. I could use other weighting option.

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Arthur Goldsipe 2023-10-31

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2040761-how-to-solve-this-error#answer_1344206

(Copying answer from comments.) The problem was that one value of DV1 was 0, so the resulting weight was infinite.

One thing I would add to the above answer is that it often doesn't make sense to use this value of 0 during fitting, since the initial conditions of the model are usually guaranteed to provide a value of 0. So you could replace the value of 0 with NaN to indicate that there is no measured data for fitting.