Is this right? Comparing age and number of puppies

Hi there (Hi) I'm doing a simple study looking at the effect of maternal age on litter size in a group of dogs. Age has been recorded in years (1-8) along with litter size (88 litters, each ranging from 1 to 8 puppies). Now someone told me that due to the large number of litter size categories that I should treat them as continuous and to use spearman's rank to analyse. Is this correct or is there a better way? I'd really appreciate any advice as unfortunately stats isn't my strong point at all and it's driving me crazy. Cheers! x

Re: Is this right? Comparing age and number of puppies

Hey vetsky.

I would not treat this variable as continuous in my opinion. Eight categories is not that much.

You should probably consider a Poisson distribution regression model which is a Generalized Linear Model (GLM) since you are dealing with count data. The nature of the regression though will depend on the nature of the data both collected as well as the domain knowledge of that data.

You can do this kind of calculation in R or SAS (R is free and open source). In R you use the glm() command to do this kind of computation after you read in the data.

What the glm function does in a nutshell is it will fit a model between age and litter size and fit an "average" line through the data describing litter size as a function of age. You can make the model more complex by adding more terms and if there is some kind of curvy or complex relationship then you add more terms to the model. It will look like a y = f(x) kind of model in the end.

What you do after this is look at the coefficients of the curve (for example, in y = mx + b if m > 0 then it means that litter size increases with age but if m < 0 then it means that litter size decreases with age), and look at what the effects are. If you get curves in the relationship, then you need to take this into account.

If you don't have a strong stats background then this can get ugly, but the idea behind the methods is very simple: fit y against x using some model that makes sense given the data and situation and look at how good the fit is and what the coefficients tell us about the relationship between y and x.