I use this address field (or the common name recorded in the call record, if the address field is blank) to calculate counts of visits to an address.
I use the OFFENSE_DATE to determine the date at which the visit to the site occoured. I calculate counts of visits based on the number of time this date field is observed on a particular date. It is important to note here that common sense suggests that the choice of the field name is for administrative purposes and not intended to be interpreted as a legal offense.
I sort all count data from high to low, this converting count fields as ordered, discrete variables.
Preparing time series data
I created time series data at the 1 day level for the observed variable.
Exploratory data analysis
Exploratory data analysis involved (a) generating descriptive statistics of the dataset. (b) Time series visualisation of the observed variable across the full dataset. (c) A simple X-Y plot of locations (the independent variable, ordered from high to low) against the count of visits attributable to the location. I also performed simple ratio analysis of the service volume information, by the locations that were attributed to 10 or more calls per day.
Model discovery and fitting
I attempted to find the appropriate stochastic model iteratively fitting the service volume data to the distribution and observing result of the ChiSquare test.
I then attempted to fit the data against a polynomial and reported the result.
Longitudinal analysis
I performed longitudinal data analysis, yearwise, with a focus on locations. I selected the top “high needs” locations that were common in each of the ten years.
Visualisation
I geocoded the locations, by (unique) street addresses and plotted the result geospatially.
Results
Time series observation(s)