New to forum having trouble with stats for research project

#1
Hi,

I was wondering if anyone could help me? I'm trying to figure out what stats I need to use to help me answer a question as part of the research project Im doing. I'm trying to work out if the prevelance of lesions in common dolphin increase or decrease over a four year period from 2002-2005. I was initally told to do a t-test but im not sure this is correct I have tryed a ANOVA but still unsure this is correct. Any thoughts??
 

jpkelley

TS Contributor
#2
Hi Kristy86...can you give us a bit of background about the project? Is this for a published report or paper, or is this for an assignment? Depending on your expertise and the assignment, there may be different suggestions for an approach. Also, how is your prevalence data structured? Typical data types for prevalence (e.g. proportion of dolphins > 1lesion) are proportion or counts (e.g. number of lesions observed on each dolphin)? In the former, you might only be dealing with four data points --one for each year. In the latter, you're dealing with a distribution that is bounded (>0) and perhaps heavily populated by many zeroes (i.e. many dolphins in the populations were observed with no lesions). Such distributions can cause a bit of havoc for finding an appropriate model (ANOVA or t-test) for your data. Anyway, looking forward to hearing more about your cool data set...
 
#3
I am looking at physical deformities, and lesions present on common dolphin within the Hauraki Gulf in New Zealand. It is for a final project as part of my post grad. I have worked out the percentage of lesions present across the population for each month as well as the percentage across each year from 2002-2005. If a lesion or lesions were present in the photograph this was listed as 1. The number of photographs taken per encounter were sorted into replicates (those with the same animal present) and those able to be used and those unable to be used. The number of photos able to be used per month vs the number of photos with lesions present was used to work out the percentage of animals with lesions present per month and per year. There is no data available on the exact number of dolphins present within the area present at this time.
 

jpkelley

TS Contributor
#4
Those like like conservative measures of lesion prevalence. Sounds good. It's great you have multiple pics per "survey" (or single time point). I assume you have sampled appropriately so that double-counts of individuals are minimized (either by sampling in spatially distinct regions on different days, etc.). I would recommend not clumping your data prematurely (i.e. by year), and this will artificially reduce your sample size. That is, it won't take full advantage of all the hard field work that was done. You're in the realm of binomial regression, maybe a binomial GLM; one way to do this is with R.