In article <[EMAIL PROTECTED]>, Xinmiao <[EMAIL PROTECTED]> wrote: >Thank you all very much for your kind replies and valuable comments. I >realized that I should have stated my question clearer - next time I >will :) I'm trying to design an experiment in which >behavioral/neuronal response v.s. stimulus strength curves are to be >measured. We know that the tuning curves are sigmoidal/linear, and my >question was how I should spread out the sampling points along the >stimulus dimension, say 4 or 7 or even more? We've also known that the >error-variance, at least for neuronal responses, should be >approximately same as mean, the square root of which shouldn't differ >much between different independent variable (stimulus strength). >Having read all your suggestions, my impression is that in this case, >more sampling may not be terribly beneficial in terms of accurately >plot the position and shape of the tuning curve. On the other hand, >more replicates, especially at one or two data points, will help us a >lot in another analysis. I may try to run a simulation that's directly >pertinent to our experiment and see what that tells me.
The problem of best experimental design for non-linear functions is rather difficult, but it still is the case that a design with few points will be optimal. For linear regression, the points can be determined more easily, but for other forms, which it seems you have, the choice of points may depend on the parameters. You need to discuss this in detail with a good mathematical statistician who is knowledgeable about this type of experimental design problem, using your exact model. -- This address is for information only. I do not claim that these views are those of the Statistics Department or of Purdue University. Herman Rubin, Department of Statistics, Purdue University [EMAIL PROTECTED] Phone: (765)494-6054 FAX: (765)494-0558 . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
