Standard error about the gradient

[Ian]

Hi there,

Before I go on, I've been using DPlot since 2004. The last time I posted on this forum was 2007 about error bars (iirc). In nearly 6 years working full time in academic chemistry I have never wanted or found a need to use any other graphing software. Quite simply, I love it.

I have hit a problem today though and I'm sure there must be a solution, I just can't find it!

After plotting several points and including their standard deviations, I would like to find the gradient of the best line fit with standard error values quoted for the standard deviations about the points. Using the curve fit only the original data is measured and the SDs are ignored.

Any clues?

Cheers,

Ian

[DPlotAdmin]

I recently had another user ask for this - just haven't had time to plug it in yet. What will probably be done is to add a second curve list in the "More Curve Fits" dialog from which you can pick the standard deviation values.

[Ian]

Probably a good shout.

It should be pretty easy to do I guess. Assuming the SD around each point if the same for +ve and -ve values, couldn't DPlot just draw two new lines based on all the SD minimums and SD maximums, take their gradients and they should be exactly the same distance apart from the original gradient shouldn't they?

Ian

[DPlotAdmin]

I might have misunderstood what you wanted. The curve fitting routine I'm using allows you to specify individual standard deviations for each point, and the points with lower standard deviation are weighted heavier for the solution. (But that feature isn't currently used.)

If you just want to show new curves with a fixed error, I need to add another new control (which is no problem). But you can do that now fairly easily by making a copy of the fit (Edit>Copy>Data Values) then Edit>Operate on Y with Y=Y+error. Repeat for Y=Y-error.

[Ian]

Hi David,

Here is an example of what I'm talking about:



i.e. two 'imaginary' gradients are determined using all upper error values and all lower error values with this giving you the standard deviation about the gradient.

Regards,

Ian