[R-sig-ME] question regarding predictor variable
burwood70 at gmail.com
Mon Mar 12 03:35:29 CET 2018
You have conditioned on total of fish measured in each length bin across the
two tow type so you could analyse total catch as a loglinear GLMM
independently of the analysis of proportion in 10 min tows relative to 15
min tows. I gather you want to jointly analyse proportions and total catch
because of any possible dependency (e.g. indicated by a positive
correlation) between the pair random effects from each analysis. You could
estimate both sets of random effects and plot them one against the other to
see if there is a significant correlation. If there is its not easy to see
how you could do the appropriate MV GLMM analysis because of the different
lengths of the two DVs.
"Sally A. Roman" <saroman at vims.edu <mailto:saroman at vims.edu> > wrote:
I have a dataset I am working with to model the proportion of fish caught at
length. I have paired observations (n=96). The paired observations consist
of a 10 minute tow and a 15 minute tow. Data collected are length
measurements in 1 mm intervals, number of fish at length, and total catch
(number of baskets) for each tow.
The traditional approach to model this type of data is to use a logistic
regression to model the proportion caught at length in the 10 min to the
total catch at length in both tows in a pair and then have the pair as the
random effect. Traditional fixed effects are length (L) and length^2 (L2).
I would like to include total catch in the model, but am struggling with how
to include the variable because there is a total catch record for each tow.
I was hoping for some guidance on if it is appropriate to include total
catch for both tows or for one of the tows, if total catch between both tows
is correlated as a continuous variable.
Dr Steven G. Candy
SCANDY STATISTICAL MODELLING PTY LTD
(ABN: 83 601 268 419)
70 Burwood Drive
Blackmans Bay, TASMANIA, Australia 7052
Mobile: (61) 0439284983
[[alternative HTML version deleted]]
More information about the R-sig-mixed-models