0
$\begingroup$

I have many plant populations of the same species sampled (around 100). What is measured is an allele on one gene, that can be A,B or C. This is what I would like to use as Y variable. Then I have the yearly rain, the altitude and the the mean temperature.

I was hoping to do something similar to an ANOVA, with allele = rain + altitude + temperature but I did not find a way to do it. Ideas?

$\endgroup$
3
  • 1
    $\begingroup$ Read up on multinomial logit and its alternatives. This would not typically be regarded as multivariate analysis, so I have removed that tag. $\endgroup$
    – Nick Cox
    Dec 17, 2014 at 15:04
  • $\begingroup$ I would never expect allele frequencies to depend just linearly on rain or altitude or temperature, but there is an enormous literature on relating biological abundance to environmental controls. More plausible models usually find a maximum for some intermediate value of any environmental predictor, so quadratics are a better first approximation. $\endgroup$
    – Nick Cox
    Dec 17, 2014 at 15:52
  • $\begingroup$ Thanx. I think I am getting on the way $\endgroup$
    – Matteo Bre
    Dec 17, 2014 at 16:23

1 Answer 1

1
$\begingroup$

Your dependent variable is nominal and has three distinct levels (A,B and C); therefore, a multinomial regression would suit your needs very well. You would basically be fitting a regression whereby the log odds for any of the alleles would be modeled as a linear combination of the three predictor variables. See also here. You would basically be assessing which of the three independent variables are associated with an increase in the odds of a gene being A,B or C in comparison to your defined reference group (i.e. one of these three genes), as well as their relative contribution to the model

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.