[This was also posted to the same question on sci.stat.consult] On 25 May 2004 15:44:34 -0700, [EMAIL PROTECTED] (Xinmiao) wrote:
> Hi, > > I have the following question: If one is limited by the number of > total data points (# sampling points * # replicates), then in order to > get a good curve fitting, which is more preferable - to have fewer > sampling points and more replicates at each point, or to have more > sampling points and fewer replicates? I suspect that it is the latter, > but I don't know any statistical reference or background for that. Can > someone please kindly help me out here? I'm sure that the answer depends on the error-of-measurement or "replication error." If a replication is going to give you exactly the same value as the first measurement, then there is little to gain in making second measurements. If it is different: Does a "replication" imply a set of points, or whole curve? - that is, does the experiment generate time-series. However, I recognize the problem from something concerning repeated measures. Power analysis there ends up being awkward because you seldom have good knowledge of all the variance terms. - maybe this Reply will be a head-start for someone who knows the problem for fitting curves. However, I wonder if the question is too comprehensive. "Curve-fitting" seems to me to be approximately as broad as "fitting some kind of linear or non-linear regression." -- Rich Ulrich, [EMAIL PROTECTED] http://www.pitt.edu/~wpilib/index.html . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
