Stanley Gorodenski - ASRX wrote:

<SNIP>

> > correctapplication that embodies time, time squared etc. > > David, > Thanks much for your response. > > I am new to this job. The assessor's office is employing multiplicative > models to predict full cash value. This is supposed to be state of the > art stuff in the area of assessing. As part of preparing the data for > developing a model, what is done here is to do a scatterplot of the > data, and then break it up visually into segments of approximately > similar rate changes, and then straight line the rate changes. The data > is then adusted form these straight line approximations to bring all > sale prices to current levels, the assumption being that the changes in > sales prices being observed is due solely to economic market factors. I > do not have the latitude at this point to change things much, but feel > fitting a polynomial is an improvement over visual straightlining. After > I fit a polynomial I will write an spss program to get rate of change > from each data point to the next which will then be applied to the > original data to bring all sales prices to current value. > > > > > What you need to do is to characterize the autoprojective > > relationship in your time series and employ ARIMA structure > > and incorporate when applicable Level Shifts, Local Time > > Trends , Seasonal Pulses and Pulse Variables ... being > > sensitive to validating the time invariance of the > > coefficients and the homogeneity of the error variance. > > I agree with you. This is what I would have preferred, but, as I > mentioned, at this point I am in training, and we have spss, not the > best software for arima models, unless things have changed since I last > looked at it. > > > > > Download for free ... http://www.autobox.com/freef.exe a > > Thanks for the URL. I do not believe I could install this software at > work, but I can at my home computer.

OK ..

> > > program called FreeFore ..and see what that gives you in > > terms of an adequate model i.e. intuitive forecasts, model > > errors (fitting errors) that are Gaussian and coefficients > > that are statistically significant. > > I attended a course in 1983 (unbelievable its been that long ago) taught > by Richard McCleary. At the end of the course he gave participants a > tape (an actual mainframe tape) of the David Pack program which I took > back to work and had installed on the mainframe by technical support. I > used this for many years before we got SAS (at my other job).

We were the distributors of the PACK PROGRAM at that time ...and he had no right whatsoever to freely distribute that program. afs charged $50-$100 per copy with DJPACK receiving royalties from us. The PACK PROGRAM was the underlying software for our first offerings in this area to automate the process and we even sold a $150 add on that did just that !

> > Your URL is bringing back a faint memory. It seems a recall a company > called Autobox or something like that which was an offshoot of the > original David Pack. Is the Autobox you refer to a new face of the > original David Pack?

yES !

I had checked into Autobox's software years ago, > but it was more menu driven, and was strongly oriented toward automated > time series modeling. I preferred the old Pack tape version I got in > 1983 where models had to be specified, and, as archaic as it was, one > quickly made changes in lines of parameters (each line being equivalent > to a card). This went a lot quicker than having to wade through > parameter screens.

Our new offering AUTOBOX 6.0 gives you both the old feel of the PACK PROGRAM and the automation if you so desire.

If I can help .. please give me a call .. glad to chat

dave reilly afs 215-675-0652 http://www.autobox.com

> Stan > > > > > To learn more about time series analysis .. check out the new > > online university totally committed to excellence in > > forecasting and time series analysis. > > > > http://www.autobox.com/AFSUniversity/afsuFrameset.htm > > > > HTH .. If you wish to chat about this ...please call .. > > > > Dave Reilly > > Automatic Forecasting Systems > > http://www.autobox.com > > 215-675-0652 > >