X-Message-Number: 32353
From: David Stodolsky <>
Subject: Re: CI growth rate decreasing?
Date: Tue, 2 Feb 2010 16:18:22 +0100
References: <>

On 1 Feb 2010, at 11:00 AM, CryoNet wrote:

>   David's selection of data misrepresents the bigger
> picture of what has been happening. Using data for
> Cryonics Institute (CI) growth since 1990 gives:

It is almost always possible to bias an outcome by selection of data.  
However, in my original analysis all data were included, unless there  
was a statistically valid reason for not doing so. The excluded data  
was from the first years and clearly had a different slope. Attempts  
to fit the remaining data compared simple curves, such as linear and  
exponential. Exponential growth of CI was the best explanation for the  
data. Next, it was noted that the slope of the exponential changed in  
1998 and a test verified that change as statistically significant.  
Thus, using the data from 1998 forward gives us the best estimate of  
what growth should be in the Internet Age. These points are documented  
in the cited web page and in CryoNet messages around that time. So,  
the implication that data were selected to support a predetermined  
hypothesis is inappropriate.


>
> There was a growth
> spurt in the 1997-2001 period which
> is probably entirely attributable
> to the fact that CI gained a website
> and people started finding our
> organization through the internet.

My analysis showed that this 'spurt', that is, 23% yearly growth,  
continued thru 2004.

I see no reason that the effect of the website would stop in 2001,  
unless there is a fixed pool of potential sign ups. If there is, then  
it supports the atheist millionaire hypothesis. However, drawing a  
conclusion after examination of the data is not considered to be a  
valid method of hypothesis testing.


> Membership growth has been roughly
> constant since 2001, with a spurt
> in 2005, which I have attributed
> to a New York Times article (we
> had a huge spurt of growth in the
> weeks of the time of that article.

This is a fine hypothesis, but without systematic data collection  
making possible a correlation between publicity and growth, it can  
hardly be considered anything more than a post-hoc attempt to explain  
events.


>
> Selecting average growth from the
> 1998-2004 period as a predictive base
> is arbitrary, as compared to
> 2001-2007 or 1990-1996.

The selection was based upon a statistical test, which rejected the  
hypothesis that growth rates were the same before and after 1998. It  
was later noted that the website went online in that year and that  
this could explain the change.



> Needless to say, these amounts of data
> are not sufficient for statistically
> significant prognostication

The data is adequate to distinguish between alternative growth models  
of the simple type we are discussing here. The fit for the data for  
1998 thru 2009 is significantly different at the .05 level, when  
linear and exponential models are compared. The linear model for  
members accounts for .99 Adjusted RSquared, while the exponential  
model accounts for only .96. CI total membership has been growing at  
69 persons/year.

Similarly, the fit for the suspendee data for 1998 thru 2009 is  
significantly different at the .05 level when linear and exponential  
models are compared. However, the linear model accounts for only .988  
Adjusted RSquared, while the exponential accounts for .991. Thus,  
while the growth in suspendees remains exponential (when 2009 is  
included), it has dropped to 10%/year. (Here we can see the real  
problem of limited data, which is identification. When 1998 - 2008  
data is fit, the linear model is better. However, when the 2009 datum  
is added, we get a slightly better fit with an exponential model.  
Either way, the data tells the same story, decreased growth compared  
to 1998 - 2004.)

So, if we combine the results for both membership growth and suspendee  
growth, we can see decreases feeding thru the system. We would expect  
suspendee growth to lag membership growth, since in most cases people  
become members some time before they become suspendees. Therefore, the  
data is consistent with a decreasing growth rate at CI. The ALCOR data  
is problematic, since it has proceeded more in spurts (and even a  
decrease) and therefore can't be modeled with these simple curves, or  
probably at all, given the limitations of the data. However, I expect  
the same factors that influence CI growth to effect ALCOR growth.


>  Membership growth has
> been roughly linear (constant number
> of new Members each year)

As I have mentioned earlier, this is a faulty metric, since it ignores  
the size of the organization. (Calculations must be on the raw data,  
that is, the total number of members, etc., if they are to be  
statistically meaningful.) If a big corporation stated it had a  
successful year, because it had attracted 3 new people, this would be  
seen as absurd. However, if the same statement was made by a startup  
which only had 3 customers to begin with, it could be seen as valid.  
So, let's agree to use the appropriate metric, which for  
organizational growth is the percentage increase per year.


There is nothing new about these conclusions. There has been talk of  
'stagnation' for some time and the recent 'teens and 20s' meeting was  
a recognition that there has been a drop in interest. I merely show  
what the situation looks like from a quantitative standpoint. My  
figures were conservative. Our best estimate of growth in the Internet  
age is 23%:

Using the 1998-2004 rate of 23%/year):

163 suspendees predicted by 2010.1.1
85 suspendees actual by 2010.1.1
(Ten patients received from the American Cryonics Society in 2004 are  
not included. Thus 58 suspendees on 2005.1.1 and 85 suspendees on  
2010.1.1)

1340 total members predicted by 2010.1.1
833 total members actual by 2010.1.1
(476 members on 20041.1)

Growth in suspendees is only 52% of expected.
Growth in members is only 62% of expected.

We are certainly not heading toward the originally predicted five  
million CI members in 2044. The linear growth predicted above means  
only a few thousand CI members in 2044.


If we review the entire history of CI growth, it is clear that the  
best fit is exponential. This is what we would expect to see, if each  
member attracted a new member with a fixed probability. The fact that  
people often recruit family members to sign up, is one example of this  
dynamic. So, the current trend toward linear growth represents a  
fundamental change in the outlook for cryonics. I have argued  
elsewhere that the movement may face a political crisis in which  
political power will be the deciding factor and therefore numbers will  
be crucial. If this is correct, then the future of all cryonicists is  
at risk, if this new growth pattern persists.



dss


David Stodolsky
  Skype: davidstodolsky

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=32353