Regression line data points and replicates

Question

Hi,

More of a general statistics question than JMP per se but I thought worthwhile to get some opinions.

I'm looking at doing some MDL (Method Detection Limit) by spiking samples at different levels. The data from the line allows me to make a regression and I use the data from the line to calculate my MDL and the linearity of my data. It is not possible to do more than12 points (though I would love to)

Up till now, I have been doing 4 levels in triplicate - 12 points (say 0, 5, 10, 15). However it not more beneficial to do 6 levels in duplicate?  I would think that you get more data on the regression and you can still get standard deviation with respect to the line.

Thanks

Peter_Bartell · Answer

In addition to my colleague Mark Bailey's great advice...if you have JMP version 13, you might want to check the Evaluate Design capability which is new to JMP 13. This new capability can answer lots of design comparison type questions.

Mark_Bailey · Answer

You could use JMP Custom Desgn for this problem. You have a single response and a single continuous factor. Enter the regression model for the curve. If the curve is linear then the default model suffices. If there is some curvature, then add a second order power. Enter 12 for the user defined number of runs. The replication that results is helpful but more concentrations will not. You only need two or three concentrations. This design is more informative than the original design (four levels) or the proposed alternative design (six levels).