I am working to create a model to predict ground nesting bee emergence (response variable) based on accumulated soil temperature (explanatory variable). I have data across 6 sites and 2 years. The response variable is defined as count of bee holes. Count data was collected every two days beginning prior to emergence to determine rate of emergence. The explanatory is defined as accumulated soil temperature. Temperature sensors recorded daily soil temperature beginning in April. From this data accumulated soil temperature was calculated by taking average soil temperature - base temperature (17C). This data yield the rate of accumulated soil degrees. How do you take two rates (explanatory & response) and creating a single linear regression?
Hello Greta,
Since the DV is a count (discrete, not continuous) variable and you have relatively few independent cases, you might try applying either a Poisson or negative binomial regression model.
Good luck with your work!
I agree with David that a count model such as Poisson or negative binomial is likely. These sort of data are often over dispersed so overdispersed Poisson (also called quasipoisson) or negative binomial regression are likely candidates. A further issue is the multiple measurements per site which might suggest a multilevel model. The multilevel model allows time varying covariates so you could add temperature or other predictors at each time point and site.
I think if your DV is no more a count its a proportion (rate) so not sure about Poisson Regression, but if u model just the count its OK!
Also there is inherent ordering in your rates?
And then there is time series nature of your data!
so need to check well before analysis.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA