JREF Challenge Statistics
- Thread starter T'ai Chi
- Start date
-
- Tags
- challenge statistics
Ok, so you are looking to infer. In a self-selected sample, any trend you find is utterly meaningless, and you should know this.
The rest has been responded to at length.
For those who are wondering...
Any numbers we are after are chosen for a reason. Suppose we want to measure somebody's size, or their speed. How big is this person? Well, it rather depends--are you planning on buying him pants? Fitting him in a car seat? Firing him through the air in a trebuchet? The use we are planning on putting the numbers to will determine what sort of numbers we gather in the first place. Numbers which are approprate for one use may be completely useless for another. How fast is this person? Again, it depends--in a sprint? A mile? A marathon? Running backward? Numbers taken to answer one question do not answer the other questions.
"Why do you want to know how big that person is?" "So I can know how big he is." This is TC's answer, and it is easily seen to be meaningless. When we do get an inkling of the sorts of questions he wants to be able to answer (at least according to his website), we see they cannot be answered using the dataset he wishes to use.
Nope. It is not inferring to see the patterns in the data as was stated in
If you understand inferring, it is making a statement about the population based on data from the sample. The above is just looking at information from the sample.
Trends in the sample are not "utterly meaningless" as you erroneously claim. For example, the number of fatalities at a particular intersection from month to month are quite not "utterly meaningless".
Seeing development in gender patterns in the data is not making a statement about the gender of those attending the seances??
Am I missing something here?
That is trivially true.
It depends on what use, sure. The use in this case would be to get an understanding of the tests done by skeptical organizations that are statistical in nature. It may work for you to know that dowsing is the most tested claim, for example. Others need an actual number to increase their knowledge.
If a person wants to know how many tests per year are done, they need to see the number of tests per year that were done. You claim it is "meaningless" (which really means "meaningless to Mercutio"), but I wonder how you'd go about answering the question of how many tests per year were done without that number.
But why do you need to know how many tests per year that are done?
To see the number? That's it? You just want to look at a number.
And you want JREF to pay for that?
This is insane.
What sort of "understanding"? There are too many possible reasons that X would be tested more or less frequently in a self-selected sample. Knowing the raw numbers on this adds nothing.
You didn't understand. That's ok.
H'ethetheth
fishy rocket scientist
Second.
ETA: Though you did make me think about statistics, which is a good thing.
Last edited:
Sorry, I missed this post before.
It depends, of course, on how (and "how" will depend on "why") the data were taken in the first place. These data are not the equivalent of "number of fatalities at a particular intersection"; as a self-selected sample, this is closer to "number of accidents you and your friends happened to see while passing a particular intersection over the course of a few months." If you noticed more, does that mean that there were more? No; it could easily be an artifact of the sampling process. You have no controls in place to assure random sampling, no controls in place to assure systematic operationalization of what constitutes "an accident". If you do find a "trend", it is meaningless, in that you have no way of knowing if it says something about the intersection or the observers!
If only some of the fatalities at that intersection are reported each month, and you have no control over what percentage are or are not reported, of what use is the "number of fatalities at a particular intersection"?
Gr8wight
red-shirted crewman
- Joined
- Jul 5, 2004
- Messages
- 1,661
Originally Posted by jj, not necessarily in this thread :
Wait! T'ai Chi is Justin? Why didn't you tell me earlier? And here I've gone and completely wasted all that time reading these threads. I started ignoring Justin months ago. I swear, people, isn't there some kind of rule about posting with multiple or serial identities? Well, there should be!
Wait! T'ai Chi is Justin? Why didn't you tell me earlier? And here I've gone and completely wasted all that time reading these threads. I started ignoring Justin months ago. I swear, people, isn't there some kind of rule about posting with multiple or serial identities? Well, there should be!
There is.
When Justin tried to sign up under yet another userID, he was suspended for a week, and his posts merged together with his present userID.
nathan
Zygoticly Phased
- Joined
- Nov 22, 2004
- Messages
- 3,477
Tai Chi has not answered
Tai Chi,
you have not answered my question in response #238, namely why have you not discovered the %of female applicants yourself from the Challenge Applications forum?
Also, I gather you want to know the number of challenge tests per year. Why have you not determined this yourself, from the same forum?
Tai Chi,
you have not answered my question in response #238, namely why have you not discovered the %of female applicants yourself from the Challenge Applications forum?
Also, I gather you want to know the number of challenge tests per year. Why have you not determined this yourself, from the same forum?
Because T'ai is not interested in the answer.
No, he doesn't. Unless you totally lack any sort of understanding of statistics, and reading comprehension, and common sense.... oh, right. Carry on, then.
Steven Howard
Graduate Poster
- Joined
- Nov 16, 2004
- Messages
- 1,797
It took you seventeen days to think of that response?
Please show where I say that. (or is drkitten right?)
Do you suggest that biased data suddenly become unbiased when they are combined?
Remember your "fatalities at an intersection" example? I asked If only some of the fatalities at that intersection are reported each month, and you have no control over what percentage are or are not reported, of what use is the "number of fatalities at a particular intersection"?
Could you answer that? Take another 17 days if you need to.
You keep forgetting I am suggesting the possibility of looking into a meta-analysis ("One could also consider doing some type of meta-analysis on the data if appropriate."), among more important things like just seeing the actual data. I'm not stating that meta-analysis is definitely the way to go and it must be done no matter what. You have some odd need to pretend I am doing the latter.
Re: your undemonstrated claim that they are "biased". If they are, I'd suggest you talk to JREF since they're the ones designing biased tests according to you.
I guess you'd have to ask the people who look at such data in real life for specifics, since that data is collected in real life (http://safety.fhwa.dot.gov/facts/stats2002/index.htm, http://www.ntf.se/english/pdf/0209Killed in Sweden.pdf, and many others), and they cannot control many things either.
But how about a simple example, for you- a number might hint that perhaps a change in the road, signals, or signs is in order.