Hi Pork Chop: I am breaking out your comment from my JC Analysis to a separate thread. You make an insightful post, but I think I can be of help to clarify why my analysis may be valid, with a reasonable margin of error.
You said,
Generalizing from very limited statistical data doesn't really have any validity. It so happens that I have served in nine congregations and in those congregations I know of two cases of child molestation, in both cases the perps went to jail. I know of one case of alleged molestation that the authorities determined was bogus. However, my personal experience is no more valid statistically than Amazing's. Nine congregations is just too small a sample.
I hate sweeping generalizations, something the Organizations is notoriously guilty of using.
There are some rules to be used when engaging in statistical analysis"
Random samples: When I was in research and test engineering I would have debates with Quality Assurance engineers because my test reports would not provide a "statistical basis" for my conclusions. They failed to understand that my tests were "one time random selections" from a batch lot of components. The tests were not designed to be able to extrapolate the data to a larger number. The purpose, therefore, was to demonstrate random selection. If the component failed, then the whole batch was brought into question the same way if the component passed, and the batch lot was blessed. It is one legitimate method to test and qualify a large batch.
The law of large numbers: Polsters do not actually count very many people when they take a poll. They use demographic guildlines to get a good cross section of people from various locations, situations, sex, age, race, etc. Then they call about 1,000 randomly selected people to respond to the poll. The results are extrapolated to the population at large. Likewise, this method is used in TV rating services. As few as 1,200 homes can be used to determine the ratings and rankings during Sweeps Week held a couple of times a year. Advertizers pay millions of dollars based on these ratings. They would not risk such large sums or get wealthy were these methods questionable. These small samples are weighted to a 3%+/- accuracy, which is very good if you ask me.
What about my JC Analysis? I admit up front that I am using myself as a random sample of "one" over a 25 year period and then extrapolating this to the JW congregations in the USA as a whole, and I preface my results with the word "IF". The excellent question you ask is whether this has any real meaning? Yes it does.
When, for example, I send out a real estate mailing to 1,000 households randomly selected, I will get a certain percentage total response, of which a certain percentage will actually do business and close a transaction. I can choose any selection of 1000 homes and get the same results. It is the law of Large Numbers. The insurance industry uses the same methodology to rate drivers, life insurance rates, etc.
To provide a statistically valid data base, I would need a demographic panel across 50 states, send out mailings to "X" number of congregations and have the Elders respond. This is not feasible because active JW Elders have been instructed not to respond to polls. So, how can I make my data meaningful.
I can ask maybe 100 ex-JW Elders to chartacterize their exposure to pedophiles over a 20 year period. Whether the pedophile was sent to jail is irrelevant. Was he a molester or not? That is all I need to know. The results I get will be adequate to extrapolate across the larger number of over 11,000 JW congregations in the USA with about 1,000,000 JWs.
What about the demographics? Yes, some areas will have few to no pedophiles while other areas will have a whole basket full of molesters. The distrubution does not have to be uniform, but simply work out with the large group according to the Law of large numbers. The larger the group, the better.
So, you knew 3 pedophiles, I knew about 6, another ex-Elder knew about 1, another will know about 8 or 10, several others will know of or dealt with maybe one or two. But, spread across the large number of 1,000,000 JWs, the averages will work out.
Conservatism: Keep in mind that I used an engineering tool called 'conservatism' to extrapolate my numbers. Conservatism is used to create a margin of safety, or assure that the analysis is valid because there is a cushion to allow for the margin of error. Were I to have eliminated all my conservatisms used, the results would have been beyond staggering.
My conservative approach, according to the corrected number of congregations that Messenger posted, means that I allowed for an average of only about 1 in 4 congregations to have a molester. My experience in 9 congregations was about 1 in 2.5 congregations that have a molester. Your experience was about 1 in 3. Another poster was 4 in 5. So my analysis, with 1 in 4 is still more conservative than these, suggesting that if we got a larger sample, my results would be well within range.
So, if my analysis is equal to reality, then the numbers of molesters and victims is still staggering. If I start using the larger data suggested by various respondents, then the numbers go up from there. It is simply mindboggling.
Is my analysis Proof Positive? No, it is not. A solid random sample, properly weighted demographically, may yield a result that is only about half of mine analysis or even twice my analysis. I conservatively estimated about 400,000 victims of molestation among JWs. Let's say I was 100% in error, or twice the actual number of molesters and victims. Does then 200,000 child molestation victims make us feel any better? Suppose I am 4 times over target, then does 100,000 molestation victims make this story any better? What if I am off by a factor of 10 times too much? Does 40,000 victims of molestation make this better? Does an organization that boasts of being so clean and pure come across very clean with only a mere 40,000 children being victims of pedophiles?
Risk factors! This is like the Toxis waste issue. How much arsenic in the drinking water is safe? Tolerable? Allowable? Worth the risk? An acceptable loss to live with? The auto industry has been known, for example, to rate the probabilities of accidents involving serious injury and death against the cost of repairing known defects. Their accountants will analyze the statistical chance of accidents and resulting lawsuits against the cost of changing and repairing a safety defect and "prove" that it is cheaper to let a few people die! It is the cost of business!
But when it is your wife, child, husband, parent, or sibling that is killed or injured for life, suddenly those accounting statistics do not sound very comforting. Likewise, the Watch Tower organization in permitting even ten pedophiles to run loose when they could have prevented it which affect only 100 victims is still an amount that is reasonably unacceptable to any person. Why? Because, by allowing just a few pedophiles to run loose means that it could ALL have been reduced or eliminated.
But, you and I between us count 9 pedophiles. Waiting, I think counted 4. So we already know that there are at least 13 pedophiles among JWs in a mere short discussion of three different people from different parts of the country. This means that there is a potential with these 13 pedophiles to have anywhere from 500 to 1500 victims.
Now, the next step is to simply take a POLL on JWD and we could get enough data to make a powerful case study, and suggest that my analysis is very valid if not too conservative. - Amazing