[R-sig-eco] R-sig-ecology Digest, Vol 104, Issue 10: Test for circular data

Thiago A. Pires pires.thiagoaugusto at gmail.com
Mon Nov 21 22:22:33 CET 2016


​Hi Philip,

Thanks for your response, it was very enlightening. I studied a little
about periodic regression and tried to do the test in R. At first I have no
fixed or random effects to add, since these data are simple, just an index
over time, measured every half hour, for a day. I am posting part of the
script to know if it is correct and if the result makes sense. As I
understand it, only one  time-related variable are significant. I am trying
to read about possible interpretations of this result without much success.

>
tad.lm<-glm(tad$index~(cos(2*pi*(tad$hora/24)))+(sin(2*pi*(tad$hora/24))),
family=quasipoisson)
> summary(tad.lm)

Call:
glm(formula = tad$index ~ (cos(2 * pi * (tad$hora/24))) + (sin(2 *
    pi * (tad$hora/24))), family = quasipoisson)

Deviance Residuals:
    Min       1Q   Median       3Q      Max
-2.1060  -0.7502  -0.3460   0.3112   2.3771

Coefficients:
                            Estimate Std. Error t value Pr(>|t|)
(Intercept)                   0.3172     0.1503   2.111   0.0404 *
cos(2 * pi * (tad$hora/24))  -1.1927     0.1979  -6.027 2.85e-07 ***
sin(2 * pi * (tad$hora/24))  -0.1641     0.1697  -0.967   0.3386
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

(Dispersion parameter for quasipoisson family taken to be 1.121044)

    Null deviance: 97.400  on 47  degrees of freedom
Residual deviance: 44.898  on 45  degrees of freedom
AIC: NA

As for the part of your answer about the aggregation test, I do not think I
understood, do you have some suggestion of literature for me to try to
understand? In that case, I have to test if some samples of the index are
significantly higher than others. I have already thought of non-parametric
tests, since this index does not have normal distribution, but I still have
not found any suitable for this case.

Thank you!!

Cheers,
Thiago



Thiago A. Leão Pires
Ph.D. student in Ecology - UNICAMP
Collaborator Researcher - UNIFESP
<http://buscatextual.cnpq.br/buscatextual/visualizacv.do?id=K4298681Z0>[image:
thiagoaugustopires.wix.com/naturevisions]
<http://www.thiagoaugustopires.wix.com/naturevisions> [image: Skype ID:
thiago.augusto.pires] <+Skype+ID:+thiago.augusto.pires>
<https://www.researchgate.net/profile/Thiago_Leao-Pires>

	[[alternative HTML version deleted]]



More information about the R-sig-ecology mailing list