Functions For Correlated Non-Monotonically Increasing Data

[raydela]

David,

I wanted to take the difference between two loosely circular enclosed Polar Chart regions ("curves") generated from discrete data points (linked from an Excel Spreadsheet). I got the following error:

[my comment in brackets]

"Any Y [magnitude] terms used in the equation must represent curves with monotonically increasing X [angle] values. Curve 1 cannot be used as is."

I initially thought this was a glitch in Polar Charts until I tried the Subtract function on a couple of circles generated by DPlot (Y= Radius*(Sin(x)^2+Cos(x)^2) for X=0 to 6.29) in which it worked -- But then again, in this case the X [angle] value IS increasing as required.

In my linked data, the points are also "going around" in a circular fashion, but they don't start right a 0 degrees. Is it that I need to re-order my data Angle/Magnitude pairs to start at the smallest 1st quadrant angle in my data set and increase from there? (My current data starts "wherever" (maybe 47 degrees) and comes back around to about the same angle.
This would be somewhat inconvenient to do -- more clutter in my spreadsheet.

I understand the limitation because certain functions can have (2) or more Y values may be at a given X value, so there would be the issue of "which Y value" to use in a subtraction. However, in my example, there is no such ambiguity because data points in one curve correlate to data points in another. Would it be too difficult to make a special Subtract function for discrete data where such correlation exists?

For that matter, it would be nice to have such "Correlated Discrete Data" versions of any other function which have the same limitation.

Ray D.

[DPlotAdmin]

In my linked data, the points are also "going around" in a circular fashion, but they don't start right a 0 degrees. Is it that I need to re-order my data Angle/Magnitude pairs to start at the smallest 1st quadrant angle in my data set and increase from there?

Yes and no. Yes, the X values must increase from start to finish, but you don't really need to reorder anything. You could start at 47 and go around the circle to 407 (and repeat as needed). I realize this may not be much better if your data is also used for something else.

I understand the limitation because certain functions can have (2) or more Y values may be at a given X value, so there would be the issue of "which Y value" to use in a subtraction. However, in my example, there is no such ambiguity because data points in one curve correlate to data points in another. Would it be too difficult to make a special Subtract function for discrete data where such correlation exists?

It's not a matter of being especially difficult, just time-consuming. Instead of a few hundred points what if instead you have a few million, for example. And to make it a bit confusing, say you have 2 curves on a plot that match up like this, but 3 more that don't (but that ARE monotonically increasing). I'll give this some more thought, but it sounds like a mess.

[raydela]

You could start at 47 and go around the circle to 407 (and repeat as needed)

I missed the obvious. Thanks for pointing this out.

OK, so this isn't so bad for me (only 8 data points per curve). I could add an "if then" type check in my Excel angle cells to check its predecessor and adjust itself accordingly.


Ray