"You don't really see many papers where a single variable is used to predict anything with any kind of power. That is such a basic concept. If single variate models were accurate and/or precise at all, we would model everything with simple algebraic equations."

You don't see that many papers using few variables because it is fairly rare that such strong relationships exist. Fortunately for predicting quail in certain areas, precipitation levels alone give good estimates.
I will only make maps that have sound science (published in peer reviewed journals) behind them.


"Now I do agree that quail populations are heavily effected by precip, however I'm a little skeptical that you think this one variable can predict population density correctly 90% of the time. Can you post your data? "

Here' some data:
Correlation between abundance of Northern Bobwhite in the South Texas Plains Region and the Modified Palmer Drought Severity Index (PMDI): r=0.90, p<0.001, where r is the correlation coefficiant. Raw precipitation values are correlated at r=0.64 (p=0.002) for the same region.
The abundance of Scaled quail in three of the six regions in Texas are correlated at r=0.69; 0.75; 0.67 with PMDI, and at r=0.57; 0.66 and 0.67 with raw precipitation data.
(Bridges et al, 2001, "Differential Influence of Weather on Regional Quail Abundance in Texas, Journal of Wildlife Management 65(1): 10-18)
Also see Heffelfinger et al, 1999, Influence of Precipitation Timing and Summer Temperatures on Reproduction of Gambel's Quail, Journal of Wildlife Management 63(1): 154-161
The list of references of relevant literture is too long to post here. The literature citings in any of the two articles listed above will give you further info, if you are interested.


"The reason we use multivariate methods isn't necessarily to predict population density but to explain population trends which we can then use to predict with some small amount of statistical power. The problem with a single variable model is that is for instance you have a discrete disease event that causes your population to crash and your only looking at your precip data to explain the trend your seeing...then your model isn't even a model it's just arbitrary application of data. "

I disagree; First of all, precipitation is only one variable but it is strongly correlated with many other variables that are harder to measure. This can be called a "proxy variable". Including this type of variable with other strongly correlated variables creates many problems in statistical analysis (auto correlation etc). Therefore, using a few carefully selected variables is far superior than including all possible variables you potentially can include. Basically, including many related variables (vegetation mass AND precipitation for instance) would produce erronous results unless controlled for.

"I understand what your doing, but I am still skeptical of the usefulness of such information. Supposing there is a sybiotic relationship between quail/precip, why would you suppose any other animal would share this relationship? Secondly, most wildlife agencies have even more accurate population information then you do, so why not just consult them? I'm not sure I understand what your angle is here. "

For quail for instance, the usefulness is simply that you will know the areas that are likely to have more quail. It is simply a waste of time going to areas that aren't very likely to harbor the largest populations. If you are out scouting a lot, this is of course better information, but most people don't have time to do that.
The information wildlife agenciesshare with thepublicis good but not very detailed. Theyare not very likely to reveal exact location with large populations. They will tell the public the general areas, but never tell you exactly where to go. The maps I make are very detailed and precipitation actually vary a lot over fairly small areas.

No I agree with that. I'm not saying in any way that we should use every conceivable variable we can get our hands on. This is of course due to diminishing returns and increased complexity. There is no reason for sacrificing accuracy for precision or vice versa just to get your equation to hit every data point. I understand what your saying and I agree with your statement above. That still doesn't explain why you think you can predict anything with nothing but precip data. I tire of this. Your speaking as if you know something the rest of the wildlife management field does not and if your methods were a) accurateand b) precise enough tomeetthe standards of an agency that manages public resources than our jobs would be much much easier. Maybe in Texas precip is really that important but I doubt it. Also, your model is not able to explain other large population fluctuations from other sources so reallly it's inconclusive at the least and arbitrary at the most.

That paper also states that: "It (Palmer Modified Drought Index)was a better indicator of quail production than raw precipitation alone."

It also goes on to state that: "Quail production was more highly correlated with this drought index in more arid regions."

Now the problem here is your taking this one sentence and running with it. Your forgetting a basic precept from Logic 101: Correlation is not causation. That is why I stated back a few posts that precip may be coincidentally correct a partial amount of time. But it fails to expalin the dynamics to what I feel would be an acceptable standard. It also fails to explain a number of changes in population like a disease event as I stated above. The actual cause (in my example disease) is causing population changes while everyone is sitting around looking at their precip data and scratching their heads. This is why univariate analysis isn't used much. It's like putting on blinders.

The author also goes on to speak about the impact of anthropogenic factors which usually has quite a large impact as well.