Presentation is loading. Please wait.

Presentation is loading. Please wait.

Statistics in Criminal Justice and Criminology (CJC) By Ali Hashemian 1.

Similar presentations


Presentation on theme: "Statistics in Criminal Justice and Criminology (CJC) By Ali Hashemian 1."— Presentation transcript:

1 Statistics in Criminal Justice and Criminology (CJC) By Ali Hashemian 1

2 Statistics for Criminology and Criminal Justice By Jacinta M. (Michele) Gau 2

3 What do statistics have to do with criminal justice or criminology ? Researchers in the field of criminology and criminal justice examine a wide variety of issues pertaining to the criminal justice system and theories of offending. Included are things such as prosecutorial charging decisions, racial and gender disparities in sentencing, police use of force, drug and domestic violence courts, and recidivism. The following are examples of studies that have been conducted and published. Examples: “Are prisoners with low self-control at heightened risk of victimizing, or being victimized by, other inmates?” “How prevalent is victim precipitation in intimate partner violence?” “Against whom do police use Tasers? Why do they use them? Are Tasers effective? ” 3

4 Data Sources The Uniform Crime Reports (UCR) The Federal Bureau of Investigation (FBI) collects annual data on crimes reported to police agencies nationwide Crimes are sorted into eight index offenses: homicide, rape, robbery, aggravated assault, burglary, larceny-theft, motor vehicle theft, and arson The National Crime Victimization Survey (NCVS) The Bureau of the Census conducts the periodic NCVS under the care of BJS to estimate the number of criminal incidents that happen per year and to collect information about crime victims Multistage cluster sampling is used to select a random sample of households, and each member of that household who is 12 or older is asked to participate in an interview Those who agree to be interviewed are asked over the phone or in person about any and all criminal victimizations that transpired in the 6 months prior to the interview The survey employs a rotating panel design, so respondents are called at 6- month intervals for a total of 3 years, and then new respondents are selected The Police-Public Contact Survey (PPCS) The Bureau of Justice Statistics conducts the PPCS periodically as a supplement to the National Crime Victimization Survey (NCVS) Interviews are conducted in English only. NCVS respondents aged 16 and older are asked about recent experiences they may have had with police. Variables include respondent demographics, the reason for respondents’ most recent contact with police, whether the police used or threatened force against the respondents, the number of officers present at the scene, etc. This data set is used by BJS statisticians to estimate the total number of police-citizen contacts that take place each year, and it is used by researchers to study suspect, officer, and situational characteristics of police-public contacts 4

5 Levels of Measurement Categorical and Continuous Variables A categorical variable that classifies people or objects into groups. There are two types: nominal and ordinal The nominal level of measurement is the most rudimentary of all the levels. It is the least descriptive and sometimes the least informative. Race is an example of a nominal-level variable A ordinal variable is classification that places people or objects into different groups according to a particular characteristic that can be ranked in terms of quantity A continuous variable is a variable that numerically measures the presence of a particular characteristic. There are two types: interval and ratio Interval Variable: A quantitative variable that numerically measures the extent to which a particular characteristic is present or absent and does not have a true zero point Example: Age Ratio Variable: A quantitative variable that numerically measures the extent to which a particular characteristic is present or absent and has a true zero point 5

6 Levels of Measurement Dependent and Independent Variables Dependent variables are the empirical events that a researcher is attempting to explain Example: Neighborhood or city crime rates, ex-prisoner recidivism, and judicial sentencing decisions are examples of dependent variables. Independent variables are those factors that a researcher believes might either enhance or suppress the dependent variable Example: It might be predicted, for instance, that compared to their male counterparts, female judges sentence defendants less harshly because they tend to endorse the rehabilitation philosophy of punishment and the importance of keeping families and communities intact Omitted Variables An omitted variable can explain the dependent variable as well as or even better than the independent variable does. The inadvertent exclusion of one or more important variables is called the omitted variable bias 6

7 Research Example of Omitted Variables Crack Baby Panic: A media and political frenzy propelled this topic to the top of the national agenda for a time. The allegations were that “crack mothers” were abusing the drug while pregnant and were doing irreparable damage to their unborn children. Stories of low-birthweight, neurologically impaired newborns abounded. What the people caught up in this panic overlooked, though, was the fact that women who use crack cocaine while pregnant are also likely to use drugs such as tobacco and alcohol, which are known to cause problems to fetuses. They are also more likely to have low incomes and little or no access to prenatal health care. Finally, if a woman abuses crack—or any other drug—while pregnant, she may also be at risk for mistreating her child after its birth 7

8 Organizing, Displaying, and Presenting Data The program SPSS is one of the most common data software programs in criminal justice and criminology 8

9 Organizing, Displaying, and Presenting Data Death Penalty Executions Since 1999 9

10 R Code and SPSS Reference load("all_us.rda") uslessdc = all_us[all_us$STATE_NAME != "District of Columbia", ] uslessdc$STATE_NAME = factor(uslessdc$STATE_NAME) uslessdc$CrimeRate = decades[, 1][match(uslessdc$STATE_NAME, decades$State)] # bring value of decade to map data library(ggplot2) ggplot(data = uslessdc, aes(x = x_proj, y = y_proj, group = DRAWSEQ, fill = CrimeRate)) + geom_polygon(color = "black") + ggtitle(paste("Violent Crime Rate in", names(decades[1]))) + xlab("") + ylab("") Pie Charts in SPSS R Code (ggplot) for Crime Rates Bar Charts in SPSS 10

11 Hypothesis Testing Researchers are usually interested in populations but, because populations are so large, samples must suffice instead Hypothesis testing is the process of making this generalization Examples: “Do Traffic Stops Alter People’s Likelihood of Calling the Police?” “Do Waived Juvenile Defendants’ Race/Ethnicity Affect the Sentences These Juveniles Receive?” “Do Multiple Homicide Offenders Specialize in Killing?” 11

12 Hypothesis Testing A good analogy to the logic behind hypothesis testing is the presumption of innocence in a criminal trial. At the outset of a trial, the jury must consider the defendant to be legally innocent of the crime of which she or he is accused. The “null” here is innocence and the “alternative” is guilt. If the prosecutor fails to convincingly show guilt, then the innocence assumption stands and the defendant must be acquitted. If, however, the prosecutor presents sufficient incriminating evidence, then the jury rejects the assumption of innocence and renders a guilty verdict 12

13 Hypothesis Testing: Example The Census of State and Federal Adult Correctional Facilities (CSFACF) contains information about the rate of inmate-on- inmate assaults at each facility and whether or not that facility is under court order for inadequate staffing. We might hypothesize that because staffing seems clearly related to internal institutional security, those facilities under court order for severely deficient staffing levels would have higher inmate-on-inmate assault rates than the facilities that are not under staffing orders. In other words, we want to know if there is a relationship between critically deficient correctional staffing (the independent variable) and inmate-on-inmate violence (the dependent variable) 13

14 Hypothesis Testing: Example The Census of State and Federal Adult Correctional Facilities (CSFACF) contains information about the rate of inmate-on- inmate assaults at each facility and whether or not that facility is under court order for inadequate staffing. We might hypothesize that because staffing seems clearly related to internal institutional security, those facilities under court order for severely deficient staffing levels would have higher inmate-on-inmate assault rates than the facilities that are not under staffing orders. In other words, we want to know if there is a relationship between critically deficient correctional staffing (the independent variable) and inmate-on-inmate violence (the dependent variable) 14

15 Hypothesis Testing: Example Null Hypothesis (H0): Prisons under court order for staffing-related problems have the same mean assault rate as those prisons not under court order; that is, the two means are equal and there is no staffing-assault relationship. Alternative Hypothesis (H1): Prisons under court order for staffing-related problems have a higher mean assault rate than do those prisons not under court order; that is, there is a relationship between staffing and assaults, with understaffed facilities experiencing significantly more assaults H0: μ1 = μ2 H1: μ1 > μ2, where μl = the mean number of assaults in prisons that were under court order for staffing μ2 = the mean number of assaults in prisons that 15

16 Regression Analysis Regression modeling is used frequently in criminology/criminal justice research In most criminal justice/ criminology research, however, merely finding out whether or not two variables are correlated is not sufficient. Instead, researchers want to find out whether one of the variables (the independent variable) can be used to predict the other one (the dependent variable) The most common use of regression analysis in criminal justice/criminology research is in the context of hypothesis testing. Just like the correlation coefficient r, the slope coefficient b does not itself determine whether or not the null hypothesis should be rejected. To make this determination, a five-step hypothesis test must be conducted 16

17 Regression Analysis Step 1: State the Null and Alternative Hypotheses The null hypothesis in regression is generally that there is no relationship between the IV and DV and, therefore, that the slope coefficient is zero. “The null and alternative hypotheses are, as always, phrased in terms of the population parameters. In regression, B symbolizes the population slope coefficients Step 2: Identify the Distribution and Compute the Degrees of Freedom The t distribution is the one typically used in regression. When the sample size is large, z can be used instead; however, since t can accommodate any sample size, it is more efficient to simply use that distribution in all circumstances. In bivariate regression, the degrees of freedom are calculated as df=N-2 Step 3: Identify the Critical Value and State the Decision Rule For example, with a two-tailed test, α =.05, and df = 23, tcrit = ±2.069. The decision rule states: If t*(obtained) is either 2.069, H0 will be rejected 17

18 Regression Analysis Step 4: Compute the Obtained Value of the Test Statistic Compute coefficient’s standard error, symbolized SE(b). The standard error is the standard deviation of the sampling distribution. The standard error is important because all else being equal, slope coefficients with larger standard errors are less trustworthy than those with smaller standard errors. A large standard error means that there is substantial uncertainty as to the accuracy of the sample slope coefficient b as an estimate of the population slope B. SE(b) is computed as: sx = the standard deviation of x (the IV) sy = the standard deviation of y (the DV) r = the correlation between x and y Now SE(b) can be entered into the t*(obtained) formula: 18

19 Regression Analysis Step 5: Make a Decision About the Null and State the Substantive Conclusion From example before, the decision rule stated that the null would be rejected if t*(obtained) ended up being either less than -2.069 or greater than 2.069 As with correlation, rejecting the null requires further examination of the IV-DV relationship to determine the strength and quality of that connection. In the context of regression, a rejected null indicates that the IV exerts some level of predictive power over the DV; however, it is desirable to know the magnitude of this predictive capability. A close match between these two sets of scores indicates that x does a good job predicting y, while a poor correspondence signals that x is not a useful predictor. ryŷ = the correlation between the actual and predicted values of y 19

20 Regression Analysis: Example “Is police academy performance a predictor of effectiveness on the job?” - Henson, Reyns, Klahm, and Frank (2010) The researchers sought to determine whether recruits’ performance while at the academy significantly influenced their later success as police officers. They measured success in several ways: 1. the scores new officers received on the annual evaluations conducted by those officers’ supervisors 2. the number of complaints lodged against these new officers 3. the number of commendations they earned 4. Gender 5. Race 6. Age 7. Education 20

21 Regression Analysis: Example The results from three regression models show that recruits’ civil service exam scores, physical agility exam scores, or overall academy ratings were—with only one exception—unrelated to on- the-job performance. The exception was the positive slope coefficient between overall academy ratings and evaluation scores (b =.06; p<.01). The demographic variables gender, age, education, and race, also lack limited and inconsistent relationships with the three performance measures. These results seem to indicate that the types of information and training that recruits receive is not as clearly and directly related to on-the-job performance as would be ideal. There may be a need for police agencies to revisit their academy procedures to ensure that recruits are receiving training that is current, realistic, and practical in the context in which these recruits will be working once they are out on the street Regression in SPSS R-square- The percentage amount of variation in the dependent variable that is explained by the independent variables 21

22 Text Mining The purpose of text mining is to process unstructured (textual) information, extract meaningful numeric indices from the text, and, thus, make the information contained in the text accessible to the various data mining (statistical and machine learning) algorithms Information can be extracted to derive summaries for the words contained in the documents or to compute summaries for the documents based on the words contained in them. Hence, you can analyze words, clusters of words used in documents, etc., or you could analyze documents and determine similarities between them or how they are related to other variables of interest in the data mining project. Text mining will "turn text into numbers" (meaningful indices), which can then be incorporated in other analyses such as predictive data mining projects, the application of unsupervised learning methods (clustering), etc. SAS TEXT MINER INFORMATION 22

23 Text Mining: Example According to a 2008 online victimization research, approximately 1 in 7 youth (ages 10 to 17-years-old) experience a sexual approach or solicitation by means of the Internet. In response to this growing concern, law enforcement collaborations and non-profit organizations have been formed to deal with sexual exploitation on the Internet. Twenty-five transcripts from the Perverted Justice website were carefully analyzed for the development of the dictionary. These 25 online conversations ranged from 349 to 1500 lines of text. The perpetrators span from 23 to 58 years of age, were all male, and were all convicted of sexual solicitation of minors over the Internet. Communicative desensitization was modified to include the use of slang, abbreviations, net speak, and emoticons in online conversations. The core concept underpinning entrapment is the ongoing deceptive trust that develops between victims and offenders. In online luring communications, this concept is defined as perpetrator and victim sharing personal information, information about activities, relationship details, and compliments. 23

24 Text Mining: Example Communications researchers define two primary goals for content analysis: 1.To describe the communication 2.To draw inferences about its meaning In order to perform a content analysis for Internet predation, the researchers developed a codebook and dictionary to distinguish among the various constructs defined in the luring communication theoretical model. They captured key terms and phrases that were frequently used by online sexual predators, and identified their appropriate category labels within the luring model: deceptive trust development, grooming, isolation and approach [Leatherman (2009); Olson et al. (2007)]. 24

25 The End 25


Download ppt "Statistics in Criminal Justice and Criminology (CJC) By Ali Hashemian 1."

Similar presentations


Ads by Google