Z, t, n, p


Ambassador to the humans
I highly suggest you read some sort of mathematical statistics book. You've made many posts but they are all so wrong that it would take too much time to correct them and you haven't shown that you would even be receptive to any suggestions. I've tried pointing that out to you before but you were extremely against the idea of even providing any info on your background.

I'm not trying to be antagonistic but I will say that there are very basic issues with your propositions that a rudimentary course in mathematical statistics would hopefully remedy. It's up to you to get up to snuff on the basics before proposing something that would change how basic statistics is done.

As it stands I don't respond in whole to your posts because you have most of the markings of a "crank". I'm not trying to be mean. I'm just being honest.

If you are serious about doing actual work then I highly suggest reading through some basic material or taking an introductory course in the subject you're trying to comment on.

So with that said I would love to hear your response and hopefully you'll either prove me wrong and ask for help on how to proceed or provide some nugget of insight that changes my mind. If I don't see that I will consult the rest of the contributors on the forum to see if they think a ban is in order because I don't want your posts wasting others time.


Less is more. Stay pure. Stay poor.
I get what you are trying and good for you in getting your hands dirty. The set n-value threshold for significance could only exist if every sample with increasing size had the same characteristics (e.g., means, variance). Which I wouldn't imagine to be feasibly possible. But if the point is that on average standard errors get smaller as sample sizes gets larger - that is correct.


Less is more. Stay pure. Stay poor.
@joeb33050 - I think you may be at the place where you need to start a new post with a newly formulated question. Most people would look at this thread and say, "no way am I going to read all this to understand the backdrop to his question".

I don't know what you are referring to with "bias". If an estimate is off from the super population value, the issue could be easily enough sampling variability. Then the estimate will always be off if you don't have the full population. In statistics using confidence intervals or credible intervals you can understand the potential coverage of the estimate to a certain level of probability. But if the true mean here was 5, can you recover exactly that value given samples from the following population?



Less is more. Stay pure. Stay poor.
Perhaps you are missing the idea that you are creating an estimate of the truth using a portion of the full data. I don't think that makes it a bias. It is an estimate that will converge to the truth as more information is included.