Dear Scott R Millis

I hope Christian Schuster (who originated this discussion) may give more information about data. What I could understand is he want to calculate Pearson's 'r'  between two variable amount (range 1-100) and ordinal variable size (small, mid, large mega).

As I suggested in earlier reply that one can calculate canonical correlation by breaking ordinal data in dichotomous variables for getting analogous to Pearson’s r. But it requires a lot of assumptions.

Unfortunately I have not access of book mentioned by you. I will appreciate if you can provide web links for said ideas.

With regards

Nand Kishore

From: SR Millis <srmillis@...>
Subject: Re: [Statisticians_group] Stata and non-numerical correlations
To: Statisticians_group@...
Date: Tuesday, 27 October, 2009, 12:25 AM

 

This book does an excellent job of discussing this topic:   Chen, P., & Popovich, P. (2002). Correlation: Parametric and nonparametric measures. Thousand Oaks, CA: Sage Publications.   Without knowing more about your data/study, it is difficult for me to give precise advice, but you should be able to get this variant of Pearson's r for correlating a continuous variable with an ordinal one by simply running Stata's Pearson r correlation routine on your data.   SR Millis --- On Mon, 10/26/09, Nand Kishore Singh <nk_singh1@yahoo. com> wrote: From: Nand Kishore Singh <nk_singh1@yahoo. com> Subject: Re: [Statisticians_ group] Stata and non-numerical correlations To: Statisticians_ group@yahoogroup s.co.in Date: Monday, October 26, 2009, 2:38 PM   Dear Scott R Millis Can you elaborate method of calculating Pearson's 'r' between continuous and ordinal variables without any manipulation of data? What is r(RI)? Whether result will be influenced by spacing between numbers used for representing the order? With regards Nand Kishore --- On Sat, 24/10/09, SR Millis <srmillis@yahoo. com> wrote: From: SR Millis <srmillis@yahoo. com> Subject: Re: [Statisticians_ group] Stata and non-numerical correlations To: Statisticians_ group@yahoogroup s.co.in Date: Saturday, 24 October, 2009, 11:40 PM   Correlating a continuous variable with an ordinal variable is simply a special case of Pearson's r: r(RI). Without any manipulation of the data,  r(RI) can be computed using the Pearson's r equation. SR Millis ~~~~~~~~~~~ "Kunst ist schön, macht aber viel Arbeit." Scott R Millis, PhD, ABPP (CN,CL,RP), CStat, CSci Professor & Director of Research Dept of Physical Medicine & Rehabilitation Dept of Emergency Medicine Wayne State University School of Medicine 261 Mack Blvd Detroit, MI 48201 Email: smillis@med. wayne.edu Tel: 313-993-8085 Fax: 313-966-7682 --- On Sat, 10/24/09, Christian Schuster <csvirtual@gmx. de> wrote: From: Christian Schuster <csvirtual@gmx. de> Subject: [Statisticians_ group] Stata and non-numerical correlations To: Statisticians_ group@yahoogroup s.co.in Date: Saturday, October 24, 2009, 1:36 PM   Dear members of the group, I am new to the group and happy to have joined! Currently I am writing on a thesis and have one issue with Stata where I can't find the right function. I want to see if two variables correlate. Both or one of the variables are/is not a numeric variable but a string ( or categorical, or ordinal). Example: Variable "size" can have the following content: small, mid, large, mega -> this variable is a string Variable "amount" can have numbers from 1 to 100 -> numeric variable Now, I want to see if there is a correlation (for example: "the correlation of a high amount is significant with size "mega" (at a confidence level of...)". I can't find a function in Stata. It always states that I need a numeric variable with all the functions I tried. The nptrend (Trend test across ordered groups) comes closest but still you have to have one numeric variable It would be great if someone could provide me with a quick solution. Thanks in advance!! -- Neu: GMX DSL bis 50.000 kBit/s und 200,- Euro Startguthaben! http://portal. gmx.net/de/ go/dsl02 Try the new Yahoo! India Homepage. Click here.