[R] Survival analysis

[David Winsemius]

On Feb 17, 2010, at 5:32 PM, FishR wrote:

>
> We are looking the extinction of a species of freshwater fish. The  
> logistic
> regression was derived by scoring the anecdotal descriptions of the  
> species'
> former population size (1 for a positive description of the  
> population e.g.
> abundant, and 0 for a negative description e.g. scarce) and plotting  
> this
> against time. Therefore it’s the population size relative to t=0. The
> anecdotal evidence in not regular and therefore why I used a derived  
> measure
> of the population.
>
> We then have the predictor variables temperature, oxygen and river
> modification for some of the 1800-2000 time period. Unfortunately  
> the data
> is collected in bursts e.g. for the oxygen 1923-1938 and the  
> 1954-1972, so
> the missing data will not be random.

If you had large amounts of river-level data that spanned intervals  
and ended with localized extinction events, I wonder if it might be  
possible to do interval type "survival analysis"? One of the  
formulation for coxph is to model a Surv object created with left and  
right censoring. It "looks" like Surv(time1, time2, event). If you  
have credible population counts over intervals, perhaps some sort of  
expansion of those observations could put the data in an acceptable  
form. (Unfortunately those very large gaps in the record make me a bit  
dubious about the validity of results from that effort that would  
generalize across a 200 year span.) Another formulation might be a  
properly constructed Poisson regression. At least that framework is  
easily configured to handle grouped data as might be found in surveys.

  There are also real statisticians (I'm just an old county doctor)  
reading this, and with yet a better description of the data resources  
they might be able to comment on further alternatives. Another forum  
for such a question might be the geospatial group.
>
--

David Winsemius, MD
Heritage Laboratories
West Hartford, CT