Re: How to program a probabilty plot?

Jack Tomsky wrote:
I am trying to write a little code to make a
probability plot. I don't think it will take much effort once I have the right
recipe. I think I have found the algorithm here.

http://www.itl.nist.gov/div898/handbook/eda/section3/n
ormprpl.htm

There are a couple of things I am not sure about.
The "uniform order statistic medians" are first called U(i). Then, when
defined, they are called m(i) as far as I can tell. Have they just
switched the symbol from U to m or is there something I am missing?
Apparently, I am supposed to know what the "ordered response values"
are because I can find a number of references to them via Google, but
nowhere can I find a definition. What are they? Also, I have found two
methods for making probability plots. One is as in the link above. The
other, which I can't find again, involves using some specially
scaled paper and plotting by hand. Are these two methods equivalent?
Thanks for any help.

Yes, they have switched notation on you. U(i) is the same as m(i). Usually for normal probability plots, the U(i) are chosen as the means of ordered uniforms, which are given by U(i) = i/(n+1). There are other expressions sometimes used.

The "ordered response values" are what you get when you sort the data values in increasing order. That is, the smallest is the first and the largest is the last.

As a side note on terminology, I would object to the article calling G(u) a "percent point function". The argument of G ranges from zero to one, not from zero to one hundred. A more accurate name would be a "quantile point function".

Thanks. I guess I have to get programming now.
.