I have a set of data to analyse on butterfly migration. Unfortunately my stats skills are not as good as my field skills!
Me and my peers have counted the amount of butterflies migrating in one direction between 12:00 to 13:00 pm every day for 30+ days. We have also taken into account cloud cover (as the butterfly would not fly when cloudy).
As a starting point, you can perform Time Series Modeling with your data.
Time Series Modeling with your data.
Article Linking the continental migratory cycle of the monarch butte...
Article Butterfly Monitoring Methods: The ideal and the Real World
http://nora.nerc.ac.uk/id/eprint/21064/1/N021064PP.pdf
From what I've understand time series analysis is used also to forecast. I am more interested in analysing N° of Counts over time. In addition I could look at how these are influenced by cloud cover. Maybe a GLM would work too?
What is your research question, I.E. what are you supposed to learn from your study? Best, D. Booth
There are too many statistical models to cope with such elementary problems, e.g., Time-series models based on symmetric Bingham distributions, Gaussian etc....For more detail background and solutions to such problems please read: http://users.isr.ist.utl.pt/~wurmd/Livros/school/Bishop%20-%20Pattern%20Recognition%20And%20Machine%20Learning%20-%20Springer%20%202006.pdf
Since the data is non normally distributed I was thinking to use a GLM with poisson regression. I want to know if the abundance is affected over time - this is quite obvious when you look at the data as there is a strong positive skew. also I would like to test how abundance of butterflies is affected by cloud cover (presence yes or now)
I think GLM is ok. However, using Poisson regression depends on whether mean and variance of abundance of butterflies are equal. If abundance of butterflies data is overdispersed, use GLM with negative binomial regression. However, you may test for autocorrelation in the abundance of butterflies data. This is because the data generation of abundance of butterflies may be time dependent.
Kindly go through
Article Modelling of malaria incidence in Akure, Nigeria: negative b...
Article A Dynamical and Zero-Inflated Negative Binomial Regression M...
I decided to go for a negative binomial general linear model and this is what I got:
Call:
glm.nb(formula = Individuals ~ Day + Clouds, init.theta = 579254.4144,
link = log)
Deviance Residuals:
Min 1Q Median 3Q Max
-12.034 -4.420 -2.175 0.607 12.255
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) 5.770103 0.029872 193.16
Hello Riccardo Mattea
execute warnings() and past results here.
Day - what is it?
Clouds - what is it? (yes/no)?
Hi Anton,
Day is just a time series so:
Day
1
2
3
...
50
Here are some of the warnings:
1: In theta.ml(Y, mu, sum(w), w, limit = control$maxit, trace = control$trace > ... :
iteration limit reached
2: In sqrt(1/i) : NaNs produced
3: In theta.ml(Y, mu, sum(w), w, limit = control$maxit, trace = control$trace > ... :
iteration limit reached
4: In sqrt(1/i) : NaNs produced
5: In theta.ml(Y, mu, sum(w), w, limit = control$maxit, trace = control$trace > ... :
iteration limit reached
Is it because of the presence of zeros?
Do I consider my analysis correct or change everything?
And yes, the CLOUD factor is Yes/NO
Riccardo Mattea
1: In theta.ml(Y, mu, sum(w), w, limit = control$maxit, trace = control$trace > ... :
iteration limit reached
set.seed(1234) #reproduce
n
- Badges
- Science topic