Lost Cowboy,
Some time ago I said that for real data curve fitting is not needed, as it destroys the orginal data. Later, on account what Aptus said, I agreed that the use of curve fitting on real data may well be useful in various ways. You asked about errors in real data. Yes, it is certainly possible. Aside form publication errors there is for stock prices the possibility that the price gets accidentally depressed. It happend to me!.
ASMI(Nasdaq) was trading at about $ 2.85 and I issued a partial sale via the Vortex AIM method but I neglected to enter as limit. The sale went throug at $ 1.60 or so(~1993). So, the low price, although real, was an error. If I had entered the limit the price would not have dropped(it was in a rising mode). Somebody made money on it as the stock kept rising, hitting $ 59 about 4 years later! I had already sold all of what I held at about $7(bought at between $ 0,80 and $1,25)
Stock value is determined by fools(now and then. I was one then). So, as an afterthought, the utility of using curve fitting or not on stock prices depends on the purpose we have in mind for what we are going to do with the smoothed curve.
I can foresee, Various answers. One of these is that if you are selling stock then you would want to sell at the real peak and not on the fitted curve price on that moment. The same goes for buying: the fitted curve woud not be useful. For analysis purposes the fitted curve could simplify the work.
In retrospect I think that it not usefull to formulate a standard answers as to the unitlity of curve fitting. It should be determined on a case-by-case issue.
I think I have, to some extent, avoided your question, or the point you made.
Conrad
AI
