Re: The propotion mean...

From: Richard Ulrich (Rich.Ulrich_at_comcast.net)
Date: 12/09/04 

Date: Wed, 08 Dec 2004 22:28:23 -0500

On 7 Dec 2004 23:23:31 -0800, jeff.wcchang@msa.hinet.net (Jeff Chang)
wrote:

> Dear Sir
>
> Thank you to respond my question.
> Let me explain what the "propotion data" is.

Google works better, spelling "Proportion"

> Simple said that, mean the "percentage" data.
> All the data are between 0 to 1.
> something like the "Good percentage"(yield) or "bad percentage"(defect ratio)
> If my data belong to this type, can't I use ANOVA ?

You can just about always "use ANOVA." The question is,
who is going to complain, how severely?

Do the differences in proportions look like "equal intervals"?
 - yeah, sometimes. IF all the proportions, in various subgroups,
are between 20% and 80%, the choice of model often doesn't
matter much.

BUT you won't find many statisticians who will defend the
arbitrary argument that the range 45-49% should be treated
as exactly the same "interval" as 1-5% and 95-99%.

Re-read your sources with that in mind. 

-- 
Rich Ulrich, wpilib@pitt.edu
http://www.pitt.edu/~wpilib/index.html

Relevant Pages

  • Re: Repeated measures ANOVA ?
    ... >> I am mainly interested in determining wether or not the animals show a ... >> assumptions of ANOVA. ... > a separate d.v./analysis with its own error term. ... Once the first two proportions are known for a subject, ...
    (sci.stat.consult)
  • Re: Within-subjects proportional data
    ... > Non-linear mixed model with logit link! ... proportions is a lot simpler and would be a useful starting point ... before doing the more complex analyses. ... I'd also consider a simple repeated measures ANOVA looking at the ...
    (sci.stat.consult)
  • Re: Fishers exact test for time allocation data
    ... Example 1 in the Shaffer paper you've pointed to in the past ... proportions of time spent in teaching, ... Faculty members are sorted into groups by discipline (rather than ... at the Groups x Categories F-test in the anova. ...
    (sci.stat.edu)