moni wrote:
> On 23 Mar, 12:43, Art Kendall <Arthur.Kend...@[EMAIL PROTECTED]
> wrote:
>> I am not sure I am reading your post correctly.
>> Is there reason to believe that it has a specific non-normal
>> distribution in the population e.g., is it a percentage or a reaction
>> time or . . .?
>>
>> do you have a single categorical independent variable with many
>> categories or do you have several crossed independent variables?
>>
>> What is the design you want for the multivariate analysis? The same
>> categorical variable(s)on the independent side? Are there other
>> dependent variables or do you have the same variable repeated?
>>
>> Also it helps us to understand you question if you describe your
>> variables and what their values represent.
>>
>> How many cases will you have in your data set?
>> Is the design a pure experiment with random assignment to treatment or
>> is it a quasi-experimental design?
>>
>> What questions are you using the data to seek answers for?
>>
>> Art Kendall
>> Social Research Consultants
>>
>
> the study is with 1 dependent variable (lymph nodes-quantitatve) and
> multiple independent (or not) variables (age, ***, type of surgery,
> the surgeon, the anatomopatologist and so on, and i have put tham
> type numeric with nominal measure).
> I can put the dependent variable with intervales.
> The distribuition of the the lymph nodes is not normal!!
> I want to do unifactorial and multifactorial analysis like:
>
> age (per unit increase in yr)
> 61-70 vs <60 ..........................coef.SE unifactorial:-0,09 and
> p<0,001 and multifactorial coef.SE=-1,267 com p<0,001
> 71-80 vs <60
> *** (female vs male)
>
>
> how can i do that?
So your outcome variable is the number of lymph nodes--i.e., a count,
right? In that case, you might want to take a look at Poisson
regression. You can get an overview at Wikipedia, and can no doubt find
lots of other notes etc online.
http://en.wikipedia.org/wiki/Poisson_regression
p.s. - When posting the same question to multiple groups, please
cross-post (i.e., write the message once with all the groups on the
Newsgroup: line).
--
Bruce Weaver
bweaver@[EMAIL PROTECTED]
"When all else fails, RTFM."
|
