fit data cloud

[waynecj7]

Hi,

I have a data set that is formed from merging multiple channels (ChnConcat) and then limiting those channel values to predetermined range.  The resulting data produces a cloud of data that I would like to be able to then perform different types of fits (polynomial, spline, etc).  The polynomial fit works fine with ChnRegrXYCalc, but I get an error about the data not being "Strictly Monotone Increasing" if I try to do a Non-Parametrical Spline fit.  Is there a way to make the data set strictly monotone increasing?  I have attached a sample of the data.

Thanks.

Wayne

[herosphro]

Wayne,

I think that in order to make the data set strictly monotone increasing we would end up mangling our data.  Essentially the monotone increasing means that our data set constantly goes 'up and to the right;' or a more mathematical definition; for all x & y, if x ≤ y then f(x) ≤ f(y). You can also find more complicated definitions online if required.  The data set in the example does not conform to this pattern, thus resulting in our error.  Unfortunately I think that if we edit the data to fit this definitions we will find ourselves destroying the integrity of the data and it might be meaningless. 

I wanted to mention this before I spent time working on getting your data to conform to this standard. I will post more information as I find it, but I think we should consider using other data analysis functions if possible.

[waynecj7]

Now that I know the definition of strictly monotone increasing I agree that the data would not be as meaningful if it were made monotone.  Thanks for the information.

Wayne

[Brad_Turpin]

Hi Wayne,

The Non-Parametrical Spline fit in DIAdem ANALYSIS requires only that the X channel be strictly monotone rising, meaning that every X value is larger than its predecessor.  Since your data cycles the X values, this is patently not the case.  Note that you also have quite a few X values that are #NV in the Excel file you posted, which I assume means NaN or NoValue.  These NoValues are also not allowed in the X channel for that (and several other) DIAdem function.  I went ahead and used the sort function in DIAdem ANALYSIS to rearrange your XY pairs in ascending order of X value.  Then I chopped off the XY pairs which contained a NoValue in the X channel.  At that point I was able to run the Non-Parametrical Spline.  The data set you sent happened to have no repeated X values.  If there had been repeated X values, those X values would either need to be edited slightly to differentiate them, or the Y values would need to be averaged together, since the routine requires that every X value be > the previous one, not >=.

Brad Turpin
DIAdem Product Support Engineer
National Instruments

[waynecj7]

The #NV is indeed a NoValue.  These appeared as a result of filtering out calculation results that were out of the range of reality.  I will try today to do the sort and trim operations in the ANALYSIS portion of Diadem and record these actions in a script to try for the other channels in my data set.

Thanks for the help.

Wayne