2/25/2011 Charts: Looking at the Data

In looking at the CDC maps, I was curious about the data. Some of the ranges seem extremely tight:

3.7 percentage points on the Physical Inactivety map
1.1 and 1.2 percentage points on the Diagnosed Diabetes map
3.1 and 3.6 percentage points on the Obesity map

Depending upon the colors used, and the number and type of ranges, radically different maps can be created. As a result, very different stories can be told.

I downloaded the data for each map, and, in a new spreadsheet, displayed the ADJPERCENT (Estimated percentage adults with diagnosed diabetes or risk factor in the county after adjusting for age) values for each category, as charts:




The shape/curve of the data means that ~80% of the data are within a narrow range, with the “10% tails” going to extremes.

80% of the Physically Inactive data are between 19.4% and 32.7% (the lowest 5% range from 10.1% to 19.4%, the highest 5% range from 32.7% to 43.0%).

80% of the Diagnosed Diabetes data are between 6.7% and 11.6% (the lowest 5% range from 3.7% to 6.6%, the highest 5% range from 11.6% to 15.7%).

80% of the Obesity data are between 24.6% and 33.1% (the lowest 5% range from 11.5% to 24.6%, the highest 5% range from 33.1% to 43.9%).

What that means is that “it is not inappropriate to make sweeping statements about the data”:

Between one-fifth and one-third of the country are Physically Inactive.
Between 6% and 12% of the country have Diagnosed Diabetes.
Between one-quarter and one-third of the country are Obese.

The maps show that, unfortunately, most of the counties that are high in one problem area are also high in the other two problem areas. What that says to me is that, to avoid Diagnosed Diabetes and Obesity, I should eat moderately and exercise.

2/25/2011 Inactivity & Diabetes maps from the CDC

After reading it on vacation, my wife and switched our morning newspaper from The Boston Globe to USA Today (with the Weekend Edition Wall Street Journal on Saturday). We have been very pleased.

From a “Business GIS” point of view, USA Today has been doing a very nice job covering the state-by-state release of the Census 2010 data, with articles about ‘Latinos guide growth spurt’ (in Texas), ‘Population continues to shift away from rural areas’ (in South Dakota), and ‘Hispanic surge seen in USA’s Pacific Northwest’ (in Oregon and Washington).

On Thursday, February 17, 2011, they published a page 1A article titled ‘Activity, health: By county’ with a subheading “CDC maps show diabetes, inactivity ‘very closely tied’”. Unfortunately, they did not show any maps!

A search on the “Centers for Disease Control and Prevention” website turned up the maps:

http://apps.nccd.cdc.gov/DDT_STRS2/NationalDiabetesPrevalenceEstimates.aspx?mode=PHY

Indicator = ‘Physical Inactivity’ is the first map, Indicator = ‘Diagnosed Diabetes’ is the second map:



A third map shows Percentage of Adults Who Are Obese:


“When you actually take all three maps together, it really does give you this clear picture that the Southern and Appalachian areas on all three — obesity, diabetes and inactivity — are very closely tied to each other. It lets you step back and gives the big picture,” says Ann Albright, director of the CDC’s Division of Diabetes Translation.

In a nutshell, that is what a Geographic Information System (or Computer Cartography, or Desktop Mapping) is very good at: allowing the user to step back and get "the big picture".

2/16/2011 Using Pie Charts to represent Bank Activity

Way back on December 13, 2010, I discussed "Income as a % of MSA Median". It is essential for any financial institution to create such a map of their Assessment Area. Using a Geographic Information System (or Desktop Mapping program), the institution has the ability to see their own activity on the map. By geocoding their loan applications, each loan application record is assigned to a Census Tract. You then display all of the loan applications as Pie Charts (one pie for each census tract); the size of the pie corresponds to the number of loan applications in each Census Tract (more applications = larger pie); and each slice of the pie represents the outcome of the loan application (Originated, Denied, Purchased, or Other). The legend indicates that the most loan applications in any single Census Tract is 27. The other Legend numbers are simply for visual reference: a pie one-half the large size corresponds to a count one-half of the maximum value (in this case, 13.5); and a pie one-tenth the large size corresponds to a count one-tenth of the maximum value (in this case, 2.7). The "pie chart sizes" are ment to be "visual references" allowing spatial interpretation (more over here, less over there) - spreadsheets, on the other hand, are very good at conveying exact information (there are exactly 25 loan applications in Census Tract 2054.00, and exactly 12 loan applications in Census Tract 2084.00).

This map shows the Census Tracts shaded by Income Level, with the Pie Charts representing Fictional Bank's 2010 activity in the area:


Of course the first two questions are:
Why does Fictional Bank have so few applications in the low-income areas of Lynn?
Why does the Bank have a much larger percentage of denials in the low- and moderate-income areas?

2/1/2011 new TIGER files for the 2010 Census, part 2

Last week the Census Bureau finished releasing their 2010 TIGER files (all 50 states, plus the District of Columbia and Puerto Rico). These contain the Census Tract boundary files to be used when mapping Census 2010 data.

The 2010 release is 74,022 Census Tracts, an 11.6% increase from the 66,317 Census Tracts for use with Census2000 data. Generally, “inner city” census tracts, and rural census tracts, remain unchanged, while the additional tracts come from spliting suburban census tracts (almost all population changes from 2000-to-2010 are “increases in suburban areas”).

For 2010, there are 3,221 counties (up from the 3,220 previously because two counties in Alaska were split into three). Although the count of Census Tracts remain unchanged for 1,827 counties,1,120 gained Tracts, and 247 lost Tracts.

Forsyth County in the Atlanta, GA MSA increased +462.5%, from 8 Tracts to 45 Tracts (largest percentage increase):



Los Angeles County in the Los Angeles, CA MD increased +14.2%, from 2,054 Tracts to 2,346 Tracts (+292 is the largest count increase).

Ziebach County in the South Dakota non-MSA area decreased -80.0%, from 5 Tracts to 1 Tract (largest percentage decrease).

Cuyahoga County in the Cleveland, OH MSA decreased +11.0%, from 502 Tracts to 447 Tracts (-55 is the largest count decrease).

for a complete list of all County counts and changes, please email
dennis.dixon@dixonspatialconsulting.com
subject: TIGER 2010 Counties