Post Reply 
Critical values
06-12-2017, 06:24 PM
Post: #1
Critical values
Been searching all over, but can't find anything, hope somebody can help.

Can the Prime calculator return the critical values for the PMCC test? E.g. for a sample size of 8 at a conficence level of 95%, the critical value is 0.6215, but I just can't find anyway to get the calculator to tell me that. It must use and calculate such values internally, so how do I get at it?

Thanks

Chris
Find all posts by this user
Quote this message in a reply
06-13-2017, 06:07 PM
Post: #2
RE: Critical values
I'm afraid I am not understanding what you are looking for here. Yes, critical values are calculated in the inference app for various calculations. Your terminology is unfamiliar to me and my (rusty) stat knowledge from when I last was in there implementing stuff in the stat/infrence apps.

Do you have a sample problem with complete data? I can probably point you at where to get it if I had that.

TW

Although I work for HP, the views and opinions I post here are my own.
Find all posts by this user
Quote this message in a reply
06-13-2017, 06:28 PM (This post was last modified: 06-13-2017 06:33 PM by Claudio L..)
Post: #3
RE: Critical values
(06-12-2017 06:24 PM)ChrisHP Wrote:  Been searching all over, but can't find anything, hope somebody can help.

Can the Prime calculator return the critical values for the PMCC test? E.g. for a sample size of 8 at a conficence level of 95%, the critical value is 0.6215, but I just can't find anyway to get the calculator to tell me that. It must use and calculate such values internally, so how do I get at it?

Thanks

Chris

I believe you are looking for NORMALD_CDF(), or actually 1-NORMALD_CDF(), but in terms of a normal distribution, all you need to do is compute the average and standard deviation, then the value for 95% exceeding (or below) the average is simply A-1.645*d (or A+1.645*d). (here A=average, d=standard deviation of your samples).
In such case you don't need anything special from your calculator. Only if you want to determine where that 1.645 coefficient comes from, or if you need to change the percentile confidence, you would need NORMALD_CDF.


Nevermind - I think you are looking for something entirely deifferent.
Find all posts by this user
Quote this message in a reply
06-13-2017, 06:57 PM (This post was last modified: 06-13-2017 06:58 PM by ChrisHP.)
Post: #4
RE: Critical values
I'm trying to find the critical values that are in the "Product Moment Correlation Coefficient" tables (sometimes also called Pearson Moment Correlation Coefficient).

The table holds values by number of items (n) and the confidence level (typically 0.1, 0.05, 0.01, 0,001)

example table here

Table

e.g when n is 8, at 0.05 confidence (one tailed test), the value is 0.621, and it's this value I want the calculator to give me
Find all posts by this user
Quote this message in a reply
06-13-2017, 08:57 PM (This post was last modified: 06-13-2017 08:58 PM by Claudio L..)
Post: #5
RE: Critical values
(06-13-2017 06:28 PM)Claudio L. Wrote:  
(06-12-2017 06:24 PM)ChrisHP Wrote:  Been searching all over, but can't find anything, hope somebody can help.

Can the Prime calculator return the critical values for the PMCC test? E.g. for a sample size of 8 at a conficence level of 95%, the critical value is 0.6215, but I just can't find anyway to get the calculator to tell me that. It must use and calculate such values internally, so how do I get at it?

Thanks

Chris

I believe you are looking for NORMALD_CDF(), or actually 1-NORMALD_CDF(), but in terms of a normal distribution, all you need to do is compute the average and standard deviation, then the value for 95% exceeding (or below) the average is simply A-1.645*d (or A+1.645*d). (here A=average, d=standard deviation of your samples).
In such case you don't need anything special from your calculator. Only if you want to determine where that 1.645 coefficient comes from, or if you need to change the percentile confidence, you would need NORMALD_CDF.


Nevermind - I think you are looking for something entirely different.

Actually, for Paersons you use the STUDENT_ICDF() function to get t for your confidence interval, for example for 95% one-tail and df=10 degrees of freedom (= Nsamples - 2) you do:
t=STUDENT_ICDF(df,0.05)
r=sqrt(t^2/(t^2+df))

And that's the value you'll find on most tables.
Find all posts by this user
Quote this message in a reply
06-14-2017, 12:35 PM
Post: #6
RE: Critical values
I think we are at cross purposes, but just in case we're not, can you add numbers to your example so that I get the same answer as in my example.

For a sample of 8, at 95% confidence level, the critical value in my PMCC table is 0.6215.

Try as I might I can't get that value using the STUDENT_ICDF() function, and then calculating r from the resulting t value it returns.
Find all posts by this user
Quote this message in a reply
06-14-2017, 12:48 PM (This post was last modified: 06-14-2017 12:50 PM by Didier Lachieze.)
Post: #7
RE: Critical values
Here is a small program with the instructions provided by Claudio.

PMCC(8,0.95) returns 0.6215 in FIX 4 mode which seems to be what you expect.

Code:
EXPORT PMCC(n,cl)
BEGIN
  LOCAL t,df;
  df:=n-2;
  t:=STUDENT_ICDF(df,1-cl);
  RETURN sqrt(t^2/(t^2+df));
END;
Find all posts by this user
Quote this message in a reply
06-14-2017, 01:20 PM
Post: #8
RE: Critical values
Thnx, was getting n and df out by 2. 10 / 8 is no good, 8 / 6 is perfect
Find all posts by this user
Quote this message in a reply
06-15-2017, 02:36 PM
Post: #9
RE: Critical values
If you'd really like a table as well, I'd stick that program in a copy of the Spreadsheet app program file, build a column with your numbers, and do something like =PMCC(A1,B1) in column C.

Instant table! :-)

TW

Although I work for HP, the views and opinions I post here are my own.
Find all posts by this user
Quote this message in a reply
06-15-2017, 04:31 PM
Post: #10
RE: Critical values
thnx, will take a look at that

Chris
Find all posts by this user
Quote this message in a reply
Post Reply 




User(s) browsing this thread: 1 Guest(s)