Types of Data

There are many types of data used in crime analysis. Most often are:

1. Crimes Reports

2. Arrests Reports

3. Calls for Service

1. CAD: Computer Aided Dispatch System
2. RMS: Records Managment System

Computer Aided Dispatch System:

-typically used for emergency operations such as police, fire, and ambulance

-provides "unit history"

-does not include all phone calls, only those dispatched to or initiated by police officers

-"citizen-generated" or "officer-generated"

Records Management System:

-stores police records as databases

-used for quality control evaluation, to look up individual incidents, to track cases, and to run reports.

Is primary type of data used by crime analysts.

Limitations:

1. Only represents crimes reported to the police.
2. Is Dynamic, or information is constantly updated.

Crime data is dynamic, or is constantly being updated.  Issues:

1. "Real-time" data: performing statistics on the most recent data can change if performed a period of time later.
2. All crime data has 2 dates: report written date and occurred date. Analyses should use report written date as it never changes. Downfall = it can affect the interpretation of crime stats (i.e. XMAS burglaries reported in Jan. look like crime is up in Jan.)

1. Crime Reports
2. Uniform Crime Reports (UCR)
3. National Incident Bases Reporting System (NIBRS)
4. National Crime Victimization Survey (NCVS)

Crime reports refer to individual crime incidents that are reported to the police and are contained within the agencies records management system. This type of reporting can vary depending on region, state, and police agency.

-UCR, or Uniform Crime Report, is a national standard for classifying crime uniformly.

-Disregards state laws.

-Collects data on selected crime: homicide, rape, robbery, aggravated assault, burglary, larceny, motor vehicle theft, and arson.

-Has a hierarchical coding system.

-Only represents aggregate numbers.

-Is voluntarily submitted monthly.

NIBRS, or National Incident Based Reporting System is an attempt to improve the UCR.

-requires a standard of classifying crime uniformly

-voluntarily submitted monthly to FBI

-Improvement from UCR because:

     -primary crime types are tracked

     -allows department to update with new data as is available

NCVS, or National Crime Victimization Survey is a measurement of crime that focuses on the experiences of victims.

-purpose is to avoid counting only crimes reported to the police and to have statistics independent of police activity.

-counts crimes rarely reported to police such as rape and domestic violence.

-began in 1973. Samples of 100,000 residents age 12 or older from 50,000 households.

-used mostly by social scientists

-Arrest data should be used sparingly for crime analysis.

-For some types of crime, such as driving under the influence, drug offenses, and prostitution, arrests are the only way to measure the # of incidents.

-Limitations: does not signify guilt; only represents selected offenders; there may be bias in who are arrested; a small # of individuals may account for a large # of arrests; and special enforcements will greatly influence the # of arrests.

Main issues:

-CAD data misrepresents actual crime since CAD can come in as one type of call and result in another type of incident.

-Location on the call may not be location of incident.

-Summary data download may only include one disposition - missing information.

Rate is a ratio proportion of one variable to another.

"Normalizing". Number of crimes in area divided by the population of the area.

Three types of rates: crime rate, arrest rate, and clearance rate.

Crime Rate refers to the "index crime", is based off population, and is typically reported by police agencies nad FBI through the UCR.

"Index Crime" are only those crimes that the UCR tracks.

"Index Crime" is reported per 100,000 persons.

"Index Crime" distinguishes between person and property crimes.

Arrest rate is the number of arrests per person.

-Typically broken down by age, race, sex, and other census variables.

-Used at national level to indicate crime.

Crime analysis do not regularly use arrest rates because census data can be outdated.

Clearance Rate is the percent of reported crimes resulting in an arrest or an exceptional clearance according to FBI standards.

-A crime is cleared if at least one person is arrested.

-The arrest of one individual may clear multiple crimes.

-A crime is cleared by "exceptional means" if circumstances beyond polcie control prevent arrest.

-Clearance rates underestimate the number of clearances because arrests do not always occur close in time to when the crime is reported.

Measures of Central Tendancy are important for describing a distribution of data and comparing variables.

-Limited use as most depend on having interval or ratio variables (numeric), whereas most variables in crime are ordinal or nominal.

-Measures of Central Tendancy include: mean, mode, median, range, and standard deviation.

-Measure of central tendancy.

-Is the average of a set of numbers.

-Limited use as they require ratio or interval variables (such as response time, number of crimes perbeat), while many variables in crime analysis are ordinal and nominal.

-Sensitive to outliers.

Measure of central tendancy.

Range is the span of data distribution for a particular variable. (What's the highest number and the lowest number?)

Measure of central tendancy.

Mode is the value that occurs most frequently in a distribution of a variable.

-used regularly in crime analysis as lists of descending frequency(where are the most calls for service occuring?).

-Also used in cross tabulations.

Measure of central tendancy.

Median is the score or potential score above which and below which one half of all the scores lie.

-mid-point, or 50th percentile.

-can be used when a distribution is very skewed.

-Median and percentiles are used instead of mean when distributions are skewed.

Measure of central tendancy.

Standard Deviation (SD) is a measure of disperson around the mean.

-normal distribution = 68% of cases fall within 1 SD of the mean and 95% of cases fall within 2 SD of the mean

-SD can help detect skewed data.

-SD is used to indicate the nature of distribution, but is not often included in regular reports.

-Inference is the process of taking a random sample of data to estimate the characteristics of a population.

-Measures of central tendancy can be used.

-Inference is typically not used in crime analysis as most of the data examined is a population.

Statistical significance is the term used to denote that a statistical test has been conducted and the findings are significant at some predetermined level.

-Key question: what is the probability that what we think is a relationship between two variables is really just a chance occurrence?

-Acceptable probability = 95%

-Usually displayed as a p-value with significance at the .05 (or 5%) level.

-A minimum of 35 cases is desirable.

Statistical Significance Test.

- A one-sample t-test tests whether the mean of a single variable differs from a specified constant.

-Does not indicate why or how the mean is different from the constant, only that the difference is statistically significant.

EXAMPLE: A PD wants to know whether an average number of reports taken by police officers differs from the constant.

Statistical Significance Test.

-A paired-samples t-test statistic compares the means of two variables for a single group.

-Computes the difference between values of two variables for each case within the group and tests whether the average differs from zero.

-Useful in pre-and-post measurement test; can test against self.

Statistical Significance Test.

-An independent t-test compares means for two groups of cases to determine if they are different (i.e. experimental and control groups).

-Should be randomly assigned.

EXAMPLE: To test the difference between the average response time by different types of calls for service.

-Statistical Significance Test.

-Correlation tests the level of associations of variables or the extent and direction in which two variables are related in a linear fashion.

-Note: correlation only tests linear relationships.

-2 types: Pearson's r (continuous variables) and Spearman (ordinal variables).

-values closer to -1 and +1 have a stronger correlation.

-Unlike t-tests, correlation has a sign that indicates the DIRECTION of hte relationship. Does not indicate causation.