I'm trying to test the effect of drug combination on cell growth, calculating IC50 for using linear regression seems straightforward but not always accurate. Curve-fitting is one of the well accepted methods.
JMP in the Fit Y by X platform can plot functions to lines. Generate the plot, then using the upside down red triangle, right click fit special. There you can select polynomial degree and y or x transformations ( e^x,1/x etc) A couple of different curve fits and a comparison of the R2 values should get you close. It is not a plug and play process, unfortunately, you will need to use JMP as a tool to get you to the answer. There may be a better platform for this, or a way to program Python or R to do the regression analysis for you.
Hi Shiv, IC50 functions are normally threshold-based and sigmoidal in shape, so you're correct that a linear regression of the raw data is normally not the best method.
There are many ways to describe the sigmoidal function - I've attached a couple examples in JMP files. The first one is using a probit function, following a GLM, a binomial distribution, and also converting concentrations to a log scale. The second one is through a logistic function, using the Nonlinear regression feature in JMP.
You may want to see what is the standard method in your field, e.g. in toxicology the US EPA often recommends a (4 parameter) logistic regression model to calculate LD50s.