[R] Regression on presence/absence matrix

[Ben Bolker]

  [I don't know whether you cc'd this to r-help or not, I'm cc'ing this
back]

  Without more context it's hard to say very much, and you might be
better off on the r-sig-ecology at r-project.org list  , or on
CrossValidated (http://stats.stackexchange.com), rather than the general
r-help list (this is rapidly becoming a statistical/methodological
question -- "what method should I use?" rather than "how can I do this
in R"?)

   As you indicate below, there are lots of statistical approaches for
prediction in this case -- logistic regression, penalized regression
(glmnet package), random forests, MARSS, ... figuring out which is best
in your case isn't trivial ...

  Ben Bolker

On 14-01-25 01:52 PM, Daniel Patón Domínguez wrote:
> Dear all:
> 
> In the book "Logistic Regression: An Overview Lawrence M. Healy
> Eastern Michigan University, College of Technology" I read this
> paragraph:
> 
> "Logistic regression techniques resolve inconsistencies associated
> with dichotomous dependent data and the assumptions of ordinary sum
> of squares regression methods. The independent variables that are
> used for outcome prediction may be DICHOTOMOUS, categorical or
> continuous."
> 
> I had used logistic regression in few occasions to describe presence
> of species along gradients with continuous variables but never with
> dichotomous. This caused my confusion....
> 
> The multivariate regression trees seem another possibility when you
> have many dependent variables too dichotomous. In other cases I had
> used multivariate analysis with vegan but the assumptions of
> linearity (PCA) or modal (CA) distributions are not always observed.
> In this cases Principal Curve Analysis (pcurve) or Non Metric
> Multidimensional Scaling using different distances can be used.
> However in this case I had more interested in prediction of events
> and not in determining factors in data or the general structure of
> data. This caused my inquire about this methodological question.
> 
> Regards,
>