GraphPad Curve Fitting Guide

Preparing data for nonlinear regression

Preparing data for nonlinear regression

Previous topic Next topic No expanding text in this topic  

Preparing data for nonlinear regression

Previous topic Next topic JavaScript is required for expanding text JavaScript is required for the print function Mail us feedback on this topic!  

You must create an XY data table in Prism, for use with nonlinear regression.

Follow these guidelines to enter (or preprocess) data for nonlinear regression:

Avoid linearizing transforms such as Scatchard and Lineweaver-Burke plots. Such plots are useful for displaying data but are obsolete for data analysis.

Transforming X values can be convenient, and will not change the results of regression (so long as the model is adjusted accordingly). Use Prism's Transform analysis to do this.

Don't smooth your data. You will get invalid nonlinear regression results. Fit the raw data.

Transforming Y values to change units or to subtract a baseline can be convenient, and will not substantially affect nonlinear regression. Use Prism's Transform analysis to do this.

Avoid nonlinear Y transforms (reciprocals, logs) unless you have a very good reason. Such a transform can be useful it equalizes the variances (so the scatter at all points along the curve is about the same), but should not be done just to linearize the data.

Enter raw replicates when possible, and not just mean and SD or SEM. Prism will fit the same curve either way, but there are two advantages of entering the raw data. First, you can plot every individual replicate so you see the actual data. Second, entering the individual replicates lets you choose robust nonlinear regression or automatic outlier elimination

If you have entered replicates, first graph the individual replicates rather than mean and error bar. You may want to later plot mean and error bar, but look at a graph of the raw data first.