I was curious about how the distribution of BFRO reports normalized by population density will change if we look at the data by county instead of by state.
Averaging data can sometimes lead to flaws in interpretation (like the statistician who drowned in a river with an average depth of 3 ft).
Thus, I wanted to check if population density (when aggregated for the full state) was skewing the data.
I decided to explore the data looking at only two states (Washington and Florida).
I selected those because WA was in group A and FL was in group B (of Mendoza’s paper); and I wanted to explore very different states.
All BFRO cases by county came from the BFRO website.
Population data by county came from 2010 census.
Land area in square mile came from state government statistics.
The first graph below shows the cumulative distribution of BFRO reports (normalized per county population density) for the 39 counties in Washington. The average from the counties (0.63) is much lower than the state average (6.08). The chart shows that the distribution is skewed by a few counties with high number of normalized BFRO reports.
Attached is a Washington County map so that folks can see where Skamania and Okanogan counties are located (2 of the outliers).
The second graph shows the cumulative distribution of BFRO reports (normalized per county population density) for the 67 counties in Florida. The average from the counties (0.03) is again much lower than the state average (0.88). The chart shows that the distribution is skewed by a few counties with high number of normalized BFRO reports. The Florida state average is a mathematical artifact and is not even within the scale of the counties distribution.
Attached is a Florida County map so that folks can see where Liberty and Levy counties are located (2 of the outliers).
The summary table compares the statistics of WA to FL and calculates the ratio of the averages. The county average was also calculated using population weighted average and land area weighed average basis instead of simple averaging of all counties.
Just as Mendoza had calculated beforehand using state average statistics, WA has a much higher number of BFRO reports (normalized by population density) than Florida. The state average data showed a ratio of 7 to 1. If we use the county average we get a ratio of 18. Thus, WA indeed has more reports per population density than FL by a large number.
In doing this work, I realized that there are some issues/problems with the data that might impede drawing hard conclusions when comparing data from different states. Some of these issues are:
Number of BFRO reports for each state has different time horizons. Some data collection started early in some states (60’s) and late in others. Maybe reports should be included in the analysis for a limited but consistent time period (for example 30 years)?
The amount of effort and staff to collect/investigate reports in FL might be different than in WA at different time periods. Thus, the WA and FL databases might not be representative samples.
If BF is a real creature and lives it both WA and FL, its population density might be totally different in FL than in WA because of different habitat/climate/food sources. Thus, number of BFRO reports per human population density could be lower in FL than in WA because the density of BF population is much lower than in FL.
Human population changes over time while the BFRO reports can go back to the 60’s in some states. I am not sure if this dynamic affects much the statistical inferences. For example, I took human population data from 2010 census but some WA BFRO reports include data from the 60’s.