On 5/31/2013 9:38 PM, Jane T wrote:
> Hi,
>
> I have a dataset and a quadratic aprroximation to the data. I need to
> divide the dataset by thequadratic approximation. However, the data
> passes through zero and results in singularity issues at and around zero.
>
> If I translate the dataset before doing the division then I loose the
> relationship since x./y ~= (x+N)./(y+N)
>
> I've also tried converting into polar coordinates
>
> Any ideas how I can avoid dividing by zero?
It's like the doctor/patient story--"Doc, it hurts when I do this!"
"Well, then, don't do that..." :)
Don't see it's possible iow...if the data pass thru zero and you're
concerned re: magnitudes then indeed there is a singularity if you
divide near/at the zero in the function, certainly.
Why, pray tell, do you think you needs must do this instead of using
residuals or some other measure of approximation/fit?
--![]()
> Hi,
>
> I have a dataset and a quadratic aprroximation to the data. I need to
> divide the dataset by thequadratic approximation. However, the data
> passes through zero and results in singularity issues at and around zero.
>
> If I translate the dataset before doing the division then I loose the
> relationship since x./y ~= (x+N)./(y+N)
>
> I've also tried converting into polar coordinates
>
> Any ideas how I can avoid dividing by zero?
It's like the doctor/patient story--"Doc, it hurts when I do this!"
"Well, then, don't do that..." :)
Don't see it's possible iow...if the data pass thru zero and you're
concerned re: magnitudes then indeed there is a singularity if you
divide near/at the zero in the function, certainly.
Why, pray tell, do you think you needs must do this instead of using
residuals or some other measure of approximation/fit?
--