Statistics: My confusion

Post Reply New Topic

confidence thread:

http://www.wrongplanet.net/postt228346.html

Not do I only desire answers for myself but I believe my questions and answers could be of use in helping parents raise their children on the spectrum. This is my reasoning as why I see relevance to the questions on the parents discussion.

Here it goes. On the thread of confidence that momsparky, asdmommyandasdkid, inthistogether and I were discussing the topic of statistics briefly came up. I think this is worth asking. I do not know what p values means or what the letter p symbolically represents in this case. Inthistogether, I do understand the rest of what you're stating. Let's say we have set O for the whole population of something. Let's say we have a subset X out of the set of O. X has to be a proper subset of O is what you are telling me am I correct? Proper subset means that this set contains any member of the set O but can't contain all members. Am I correct?

I do not really have a major background or understanding in statistics. I know bits here and there therefore I do not understand how the statisticians come up with some of their tenets.

Assuming we have a reasonable sample of statistical samples like you state Inthistogether and we don't obtain all of these sample but just a subset let's do a statistical analysis of the confidence levels. By the statisticians logic we would not find one that has the confidence levels of 100%. This is what you all state that they state as true or am I misunderstanding? By the very concept of not being able to find one that is 100% then would it be reasonable to state that After doing a statistical analysis of the confidence levels of a subset of statistical samples that one can be 100% confident that one will not find a statistical sample that has a 100% confidence level. If this is true do we not have a contradiction in the statisticians' logical reasoning?

If what I state is true then I have proven by contradiction that we have one counter-example to what the statisticians' logic and we do have one statistical sample that at least has a confidence level of 100% which is the examination of some statistical confidence levels.

If what I state is false then there is an example of a statistical analysis that one can find a confidence level of 100%.

Do you all now understand why do I not grasp a lot of things that people in general say, what people state makes no sense to me, and why I am ignorant on many matters?

Last edited by cubedemon6073 on 18 Apr 2013, 9:31 am, edited 1 time in total.

Most people don't understand statistics and probability, which is a good thing for the casinos and lotteries.

When I was in college P meant probability, and was between 0 en 1 (1 is 100%). I'm not sure I understand the rest of the question. Perhaps you should ask this in the Computers, Math, Science and Technology subforum, there are some people there who know a lot about math. Perhaps also link to the original thread.

I don't know that the broader points will help you. When someone does a study of something, ideally they would want the group of people in the study to be identical to the total population that they are trying to apply their findings to. The thing is, you do not know what you do not know. if you knew the exact properties of the population you are talking about, you would not need to do a study.

So you try to use a good methodology in doing your study, and hope that this sampled group reflects the population you are studying as a whole. There is intrinsically uncertainty about this. One tries to sample randomly, and use a large sample, etc to minimize issues with the study. There are many resources on the net, I am sure (I think Khan Academy has resources on statistics, if you like Khan Academy) but the part that I think is most relevant is that even following best practices your sample will never be exactly like the population you are studying, and you have to be satisfied with "good enough." Also other people will repeat your study and if they concur with you, over multiple studies (repeatability) you have better certainty, but still not 100%.

By what you state, I see where I am going wrong then. It's because I didn't understand what statistics was. Statistics are a study of a subpopulation of a total population. I had difficulty with the definition and the concept of statistics then. It is not a study of studies but a study of population and population groups. Am I on the right track now?

Yes, you are taking a sample of the general population you are trying to study and hoping it lines up with that general populations, Then you analyze the data from your sample and make conclusions that you hope fit the population.

Silly me, I understand where I was wrong now. I had the concept and definition of statistics wrong. I interpreted the definition and concept into areas I should not have. Statistics wasn't designed to analyze other statistical studies in of themselves in the way I was doing it.

You know, I could understand a lot more if other people was like you, momsparky and inthistogether and show me where my reasoning is off. What usually happens is I ask the average joe online and I'm called a moron.

Most of the Internet forums unfortunately are not really designed for education of give and take. I tend to stay off most forums, for that reason.

My brain is fried right now, my friend. I will have to read this thread when I can follow logic and think logically.

I am at a serious disadvantage in conversations like this. It appears that the part of my brain that stores concepts does so in some kind of nonverbal format and the part of my brain that stores terms and definitions does so in a verbal format, but it resembles a matching test. IOW, I know statistical terms and statistical definitions, but I am not sure which ones go with which, and I know the concepts, but I do not know the "right" words and the concept is not in verbal form, so I have to somehow translate it into words and then figure out which words are the "right" ones.

I don't have the energy for all of that right know. But I did want to let you know I am interested in the conversation and in helping you. I'm just not in a frame of mind where I can right now. I only made it part way through your OP and I can't even begin to read the responses, so it is very possible someone else has already answered.

Yes, the idea of statistics (in this case) would be to study / measure a subset of a larger population, with the intent of taking the results and applying / extrapolating them to the larger population. The subset should be chosen in such a way that it is as representative as possible of the larger population with respect to whatever you want to measure, otherwise whatever you measure against the subset will not have value if you try to apply the results to a larger population.

In any case, there are statistical models, such as the standard distribution, which suggest basically that if you have a population of size X, and a representative sample from that population of size Y, then there is a correlation between whatever you find out about group Y and how accurately that would apply to group X. Generally this is called the standard deviation - the standard deviation tells you how accurate your numbers are likely to be For instance, in political polling, there may be 250 million voters. You may call up 10,000 of them and find out that 4900 of them are Democrats, and 5100 of them are Republicans. That doesn't mean that 49% of the 250 million population will be Democrats, and 51% will be Republicans. But given the sample size of 10,000 and the population size of 250 million, you can calculate a likely error range using standard deviation. That's how, for instance, these polls you see on TV will say something like "43% of likely voters say they're voting for <whomever>, and the poll has a margin of error of +/- 3 percentage points.

Getting back to choosing the subset of people correctly - in the example above, you would want to chose voters at random, rather than, say, all from a specific neighborhood or all men, or only voters over 50. You want the sample group to be as reflective of the overall group of voters as possible.