Data Analysis & Empirical Decision Making (EDM)
MPA , CMFO
City of Athens, TN
"For those of us who have
mistaken passion for a solution...
hope is not a course of action."
-Poet Buddy Wakefield
P-Values & Grouping.
A statistical method used to determine coorelation/no correlation between to data sets.
The practice of categorizing cities based on population, MSA status, proximity to the interstate, and service population.
Athens, Paris, Crossville, Sevierville, and Springfield
Bartlett, Brentwood, Franklin, Goodlettsville, and Red Bank
Sub Major City:
Morristown, Cleveland, Kingsport, Greenville, and Tullahoma
Knoxville, Chattanooga, and Murfreesboro
The Effects of Grouping
For Micropolitans a moderate and Major Cities a strong negative correlation was established between Inspections and Structure fires per 1,000 residents.
Micropolitans P= -0.50 Moderate Negative Relationship
Major Cities P= -0.98 Very Strong Negative Relationship
For Suburbs and Sub Major Cities we found only moderate correlation between Inspections and Structure fires, and in the case of Sub Major Cities a positive correlation per 1,000 residents.
Suburbs P= -0.08 Very Weak Negative Relationship
Sub Major Cities P= 0.47 Moderate Positive Relationship
If we look at the correlation with all benchmarking cities we see a strong positive relationship, P= 0.55.
The Effects of Grouping
Pearson's R for all Cities = 0.55, but when you break down the cities into their groups you see a different story.
For Micropolitans we found no relationship correlations between the Number of Sworn Officers and TIBRS A crimes per 1,000.
Micropolitans P= 0.14
For Suburbs, Sub Major and Major Cities we found a very strong correlation between Number of Sworn Officers and TIBRS A crimes per 1,000.
Suburbs P= 0.83
Major Cities P= 0.85
Sub Major Cities P= 0.58
Council member X - "Police Chief Ziegler do you know why we have so many Robberies this year?"
Police Chief Ziegler - "It's how we measure time."
We use Z Tests and T-Tests to determine when a statistic becomes Statistically Significant and when it is just more/less than the year before.
1. Establish the most statistically significant correlation.
2. Determine a Line of Best fit (may or may not be linear).
3. Set expectations for your outputs based on your inputs.
4. Review your results and determine if the Line of Best fit is matching your results.
Predicting Using a Line of Best Fit.
Please use Google Chrome to obtain the best export results.
Public - 3/29/16, 1:02 PM