"Tim, I think the significance of my post may have gone overhead.
"Every odd year = 2 years.
"Consider the upgrade cycle median year, the variance of the upgrade cycle, and then look at the graph and the surface area and consider how pushed out upgrades have been during the bust. "
No, I was making light of the unsupported claim that odd years are up years. A simple eyeballing of the graphs show this.
Looking at the first one Gottfried gave us:
home.attbi.com
I then eyeballed each year and marked it as "flat" (close to flat), "up," or "down." Those with more time can do more arcane analyses, but eyeballing is sufficient for my purposes. The results:
(I used the bookings. Using the ratio of bookings to shipments produces roughly the same results. Those with time are welcome to generate their own tables of ups and downs for the ratio, for the second graph that Gottfried referenced, and so on. I'm convinced the claim is not supported.)
1995: flat, or very close to it (up in first few months, then flat)
1996: down
1997: up
1998: down (down, then up, ending almost flat)
1999: up
2000: up (up, then down, ending almost flat)
2001: down
2002: up
2003: insufficient amount of time
Odd Years: 2 up, 1 down, 1 flat
Even Years: 2 up, 2 down
Yep, looks like a clear trend to me. Not.
One doesn't have to spend time calculating Poisson distribution chances for such runs to be seen to know that the "null hypothesis" (that there is trend, just chance) is not rejectable.
Now it may be that there are in fact deeper reasons for the ups and downs, but it sure isn't supported by looking at these graphs.
Granted, I didn't look for "candlestick formations" and "head and shoulders" trends. My numerology and astrology are not up to it.
--Tim May |
