In article <[EMAIL PROTECTED]>, Donald Burrill <[EMAIL PROTECTED]> wrote: >Why re-invent the wheel, and have to de-bug it besides? Get one of the >standard statistical analysis packages (SPSS, SAS, MINITAB, ...) and use >the multiple regression routine to fit your data to a suitable model.
>Minitab Inc. in particular offers a 30-day (I think) trial of MINITAB >software: visit <http://www.minitab.com> >On Wed, 13 Aug 2003, Top Spin wrote in part: >> I would appreciate suggestions for text books or reference books on >> exponential decay functions, probability distribution functions, and >> the like. ... <snip> >> I am trying to explore whether memory fades exponentially in a way >> that is similar to radioactive isotopes decaying, batteries >> discharging, or light bulbs burning out. I want to write some software >> to gather data and test these ideas, but I need help with the math. I >> want to fit the appropriate function to the test data and then use >> that function to predict future data points. > <snip, the rest> >An exponential decay function (IIRC) is of the form > Y = a * e^(bX) >where Y is the response variable that is thought to decay exponentially, >X is the variable (usually time) with respect to which the decay occurs, >and a and b are real numbers (in practice, rational numbers) to fit >the data. Taking (natural) logarithms, > Log(Y) = log(a) + bX >which is of the standard linear form. A simple linear regression >analysis will provide estimates of log(a) and b. Watch out for this one! If one is dealing with quantities subject to error, the error many be uncorrelated with the independent variable in the original model, but is likely to be rather correlated in the logarithmic model. Also, if it is counts, which seems to be the case here, 0 rears its ugly head. It would be better to set up your likelihood function and use maximum likelihood or something similar. -- 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/ . =================================================================
