## Please enable JavaScript to view this site.

 Troubleshooting fits of dose-response curves

## First consult the analysis checklists

If your goal is to interpolate from the standard curve, look at the analysis checklist designed for that purpose.

If your goal is to fit a model to determine parameter values, review the general analysis checklist for fitting a model.

Are the X values concentrations (or doses) or the logarithm of concentrations (or doses)? Prism offers equations for each. You can also use the Transform analysis to make a new results table where the X values are logarithms. Note that stretching an axis to a log scale is not at all the same thing. Prism's nonlinear regression fits the values it is given. Using the Format Axis dialog to stretch the X axis to a logarithmic scale does not change the values the curve fitter sees.

Is the top plateau defined by the data? If not, can it be constrained or shared? If the top plateau is not defined either by the data or by a constraint, then any fit of a dose-response curve is unlikely to be useful.

Is the bottom plateau defined by the data? If not, can it be constrained or shared?  If the bottom plateau is not defined either by the data or by a constraint, then any fit of a dose-response curve is unlikely to be useful.

Were the data normalized? To what values? If the data were normalized, consider constraining the top and bottom plateaus to be 0 and 1 (or 100) in the Constraint tab?

Is the Hill slope fixed? A Hill slope of 1.0 or -1.0 is commonly seen in many systems, but not all.

Four parameter assumes symmetry. The usual equations to fit dose response curves have four parameters (top, bottom, EC50, Hill Slope), and define symmetrical curves. You can choose an equation that adds a fifth parameter to fit asymmetrical curves.

Are you fitting the relative IC50 but expecting an absolute IC50, and so are surprised by the results.