Endurance.Net Ridecamp Archives

Susan:

I am sorry if I appeared to be slamming your peer reviewed article. I 
certainly will read it and comment if appropriate, but this will probably 
not happen until next winter, there are to many outside things going on 
right now.

Please let me try to explain some of my feelings about studies, journal 
articles, and correlation effects where controlled, confirmatory experiments 
either can not or have not been done.

First, I expect that I will find that the article will carefully define the 
population that was used in the study. This (unless I am really off base) 
will be those who chose to enter Tevis one particular year. Later in the 
article there will most likely be a paragraph explaining how the conclusions 
may not apply to other groups.

The discussion on Ridecamp was about generalizing this study to all 
endurance horses and riders and some posters started talking about how 
really big Arabian horses are most likely less sound than ordinary Arabian 
horses. Immediately we do not know that the larger horses, who obviously 
were those entered in the ride, were large Arabian horses.

My background relates mechanical/chemical engineering and the manufacturing 
of products for sale. We often had to start studying a problem by attempting 
to correlate properties of the final product with inputs. This is similar to 
what you were able to do at Tevis. While we never had the time and/or money 
to do enough experiments to track down all of the root causes we did have 
the ability to do some experiments to further understand the system. The 
eventual results of these experiments are what has led to my skepticism 
about the extrapolation of a correlation study to either cause and effect or 
to application to more that the original population. This is not a slam at 
the scientific quality of the original study, it is just a realization that 
things are not always as they seem

It is not uncommon in an initial study for an important variable not to be 
even considered. I have scaled up products to production where one of the 
important variables was not even considered until after 3 or 4 months of 
inconsistent production.

The classic one I ran into early in my working life was the effect of room 
humidity on the quality of magnetic video recording tape. A large number of 
very smart people had been working for a number of years trying to figure 
out why sometimes the quality was better than at other times. At an informal 
meeting over coffee someone said "Maybe it is the phase of the moon". After 
the laughter died down, the group decided to plot the day of the year when 
good and bad product had been produced. It turned out that late summer and 
winter in the Midwest were both times when bad product was more likely to be 
produced. It turned out that either to much or to little water in the grind 
would produce "bad" product. As soon as this was understood it was possible 
to fix the problem.

Another thing I noticed is that when a sample is relatively small, it is 
possible for the statistics to say something is significant when it really 
is not significant. I have had conversations with industrial statisticians 
about this and have never received a good explanation, but I do know it is 
true. A correlation study or even experiment can say that at a 95% 
confidence level a coefficient is significant. Usually these coefficients 
are relatively small (that is they are responsible for say less than 25% of 
the observed variation), but statistically significant. Being a stubborn 
person, and having the luxury of dealing with non biological systems, I 
often would do further experiments which would not show the effect. If I was 
smart, or more often lucky, I would also have decided to measure and control 
a new variable which turned out to shed a lot of light on the process. I 
also noticed that these statistically significant, but ultimately misleading 
results were often in interaction terms. The total weight of the horse and 
rider team is that type of interaction term. Again I do not know why, but 
that is what I have observed

To be continued....



Ed

Ed & Wendy Hauser
2994 Mittower Road
Victor, MT 59875

(406) 642-9640

ranch(at)sisuwest(dot)us 



=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=

Ridecamp is a service of Endurance Net, http://www.endurance.net.
Information, Policy, Disclaimer: http://www.endurance.net/Ridecamp
Subscribe/Unsubscribe http://www.endurance.net/ridecamp/logon.asp

Ride Long and Ride Safe!!

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=