We have started to look at the data sets on the website http://gozips.uakron.edu/~rps. I have started to accumulate data. My major question is to see if genetic variance was preserved better in either of two breeding regimes. Essentially I have an entire population, saved for 10,000+ generations, and their entire genotype. When we get to ANOVA, that will be a stat test I run ad-nauseum. Most of the tests so far seem irrelevant for my data set due to the fact that i have everybody(not a sample...EVERYBODY). Any comments are appreciated.
p.s. I'm having stat trouble and have decided to do every recipe in Julia Child's cook book.
This class was a fail. A damn fail. - Drew 2013
This comment has been removed by the author.
ReplyDeleteHey Drew, nice you got it working! =)
ReplyDeletemy 2c on your issue:
As far as I understand it, the idea of whether you have a population or a sample is more of a conceptual one than of an empirical one: it depends on the hypothesis you are aiming to tackle.
If you are only interested in verifying the differences between your two outcomes, you can consider them statistical populations and treat them accordingly. However, if you do that, you are testing the hypothesis that those two sets of data are equal (for example) - and nothing more. You can't extrapolate or generalize based on your findings. If you are interested in answering the question: "Did the breeding regime I conducted affect the outcome of this one simulation", then you don't even need to do any stats: you can just go ahead and compare the means. Or you can do Z tests or whatever to infer how much (instead of "if any") of an effect it had. I've heard there is "such thing as population-level statistics" as a discipline, but I never found anything other than Z tests...
But keep in mind that this reasoning is also valid for things that we empirically *know* that are samples. For example, you can go out and capture a bunch of male and female frogs. They are a sample of the population - you know that for a fact. But if your question is "Do these frogs that I captured (these and only these, I couldn't care less for the frog population or species or whatever) differ in size?", you could still treat them (statistically) as a population.
However, if you are interested in a broader question - say, "Does the breeding regime affect genetic variance in simulations", then you still have a sample, and not a population - a (friggin huge) sample of all possible outcomes that you could have had under such conditions. In that case, each individual (?) could be an observation, and you have a sample of the possible outcomes you could have resulting from that sort of simulation, and you can use that to infer how the breeding regime can affect the "population" (in this case, all outcomes possible - and not only the ones actually obtained - from such a simulation).
You can go even further and consider your WHOLE simulation as a single data point - such as, for example, if genetic variance is a single parameter calculated based on the whole sample. In this case your whole simulation could be treated as a single observation, and you could have a sample of simulations (which are part of a "conceptual population" of simulations).
So, boiling it down to a single sentence: whether you have a population or a sample depends not on how you got the data, but on how you are framing your question.
...or at least that's what I understood of all this. =P
And don't forget to invite your classmates to the feast! ;p
Thank you Don Rafael. I suppose that I will treat my demes as individuals and see how they change over time. Each deem will be compared to the original wild population and a measure of the remaining varaiation (# of quantitative trait loci) will be analyzed. Also, the mean fitness has been calculated for each deme and any deviation from the average fitness of the initial wild populations is an indication of the efficacy of the different breeding regimes.
ReplyDeleteAgain, your comments have helped me to better fram my questions. Hopefully more of our compatriots will post any questions soon.
Couple quesions. How will you compare your demes to wild populations? What do you mean by analyzing a measure of the remaining variation?
ReplyDeleteI will answers these questions in reverse order. 'Wild" populations for me are ones that I created (computer simulations of 100 demes, 150 individuals per deme, for 10,000 generations)based on parameters of a typical wild population that may be found in a zoo and that may be used for release after captive breeding. So I have genotypic data for every individual, wild, captive and released. The quantitative loci can be compared between these three groups to see if the number of alleles at each loci changes over time or if certain alleles are lost over time.
ReplyDeleteEssentially, captive breeding programs set out to maintain allelic variation in zoos by outcrossing individuals as much as possible, while at the same time exposing these critters to selection. The only way to effectively eliminate selection is to fix certain alleles via inbreeding, which forces an increase in homozygosity. In theory, in a two-allele model the chances of a loci being fixed for either allele is 50% under an inbreeding regime.
If my simulations show that my released individuals have less genetic variation, the assumption is that this populations will not be favored in a novel or variable environment. If the variation in my simulations for inbred lines is greater than the current standard practices of the AZA, then I expect to see fewer alleles lost to selection and an increase in long term fitness and evolvability.
Yeah science.