X-Message-Number: 29008 Date: Tue, 23 Jan 2007 17:02:59 +0300 From: Igor Artyuhov <> Subject: Re: To all fellow statistic tragics out there (CryoNet #28999... References: <> The trick is in using the quadratic trend model instead of the much more obvious exponential trend. If one used the latter, he'll obtain for membership forecasts something like (a very rough estimate!) M2024 = M2004*( M2004/M1984) = 476*476/16 = 14161 (instead of 1512.65 for quadratic trend) and M2044 = M2004*( M2004/M1984)^2 = 476^3/16^2 = 421289.75 (instead of 3378.32). As well, for patients' number we have P2024 = P2004*( P2004/P1984) = 68^2/2 = 2312 (instead of 211.415) and P2044 = P2004*( P2004/P1984)^2 = 68^3/4 = 78608 (instead of 461.802). It should be kept in mind that any such projections are applicable only for a reasonably short forecast time in any case, it shouldn't be comparable with the base period (in our case it is 2004-1975=29 years). So, any extrapolation for as much as 2050-2024=26 years is absolutely unsupported. We can use it to forecast numbers for up to, say, 2010; the exponential trend model will provide more reasonable forecasts than the quadratic trend, but both outcomes will be close. And the last remark: There is no such thing as correlation between values unless both of them are stochastic ones. That is, you can formally calculate the "correlation" between CI membership and, say, the position of Saturn on the sky for last ~10 years and it will be close to 1! but it would give us no usable information at all. Happy eternity, Igor Artyuhov. | Message #28999 | From: "marta sandberg" <> | Subject: To all fellow statistic tragics out there | Date: Mon, 22 Jan 2007 11:55:31 +0800 | | I am a sucker for statistics and I came across this site when browsing. | http://cryonics.home.att.net/CI/ciprojection.html | I believe it was compiled by Brent Fox. | If there is any fellow statistic tragics out there, give it a go. Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=29008