关键词 > STAT350/353
Qual Exam 2022: 350 & 353
发布时间：2023-08-25
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit
Qual Exam 2022: 350 & 353
An Overview of the Problem
In the United States, gun violence in K-12 schools has grown rapidly over the past two decades. For example, the mass shooting at Uvalde Elementary in Texas (2022) received a large degree of media attention. While the scale of this event was extreme, however, gun violence of smaller scales is more common .
As gun violence increases, researchers and policymakers continue to search for solutions. These include ideas like increasing monitoring of social and mental health of students, using metal detectors, stationing police in schools, among others. This question - What can we do to reduce gun violence? - provides the background for this exam.
The SSOCS Data
“The School Survey on Crime and Safety (SSOCS) — a nationally representative survey of U.S. K–12 public schools — is managed by the National Center for Education Statistics (NCES), an agency within the U.S. Department of Education’s Institute of Education Sciences. SSOCS collects detailed information from public schools on the incidence, frequency, seriousness, and nature of violence affecting students and school personnel. SSOCS also collects information on the programs, practices, and policies that schools have in place to prevent and reduce crime. Data from this collection can be used to examine the relationship between school characteristics and violent crimes in regular public primary, middle, high, and combined schools.”
All of the information that you need to understand this data is provided. This includes:
• SSOCS(2017-2018)Data.csv : The data
• ssocs codebook.pdf : The code book
Notice that in the code book, the Appendix A includes the actual survey and that Appendix B includes a list of all the variable names and definitions. Further information on the creation of composite variables (those ending in “18”) can be found in Chapter 5.
(Throughout, pay particular attention to data with values of “-1”. These are purposeful skips and in many (but not all) cases may need to be re-coded to “0”.)
This Exam
The purpose of this exam is to test your ability to put to use all that you have learned in STAT 350 and 353 in the context of real data, with a real question. This involves combining your understanding of regression concepts and theory with the implementation of these in code and clear interpretation to a lay audience. Be sure to convey what the results tell you, what assumptions they require, and any limitations in your results.
For this exam, we will focus in particular on two outcomes:
• INCID18 : total incidents of any crime
• DISFIRE18 : total use of firearm or explosive
To simply the analysis, you can assume the schools as a simple random sample, and ignore the sampling weights / jackknife replicates.
Finally, a strong exam is one that is judicious in what is presented (you can put materials in an Appendix), that explains the decisions and assumptions that were made and why, that explains the how the results should be interpreted, and that is clear in any limitations.
Part I. Testing Hypotheses
As stated above, researchers and policymakers have hypothesized and enacted a variety of policies meant to reduce crimes and gun violence in schools. In particular, they often argue that schools should include security guards in order to reduce crime and gun violence.
For this part, answer the following questions:
1. After exploring the two outcomes (INCID18 and DISFIRE18) determine what type of regression model is appropriate for each (e.g., OLS). Explain your choice.
2. Are the presence of security guards (SEC_FT18 and SEC_PT18) associated with reductions in crime (INCID18) and gun violence (DISFIRE18)? Interpret the effects clearly in language that a non- statistician could understand.
3. To what extent do these effects differ in urban schools versus non-urban schools?
4. Do your analyses suggest that policymakers are correct that security guards reduce crime and gun violence? If so, explain why. If not, conduct additional analyses (using regression) that allow you to evaluate their claim and interpret your results.
Part II. Predicting Crime
Other researchers and policymakers would like to develop a model to predict crime (INCID18) based upon observable school characteristics. Their idea is that they could first predict schools that have a lot of crime and then put in place interventions that could reduce such crime.
For this part, perform the following tasks.
1. For your first model, use variables C0532, C0534, C0536, C0538, C0560, C0562, C0568, FR_LVEL, FR_URBAN, and FR_SIZE as predictor variables. (In addition to Appendix B, you can find more detailed explanation for the variables C0532 to C0568 on pages 80-81 of the code book, and the three variables FR_LVEL, FR_URBAN, and FR_SIZE on page 172). Perform diagnostic analysis and draw your conclusion. Is there any further action you would like to do?
2. Develop and implement an approach to build the best model possible that predicts the total number of crimes (incidents, INCID18). In addition to the variables mentioned in the previous problem, what other variables would you like to consider and why? What is your final model and why do you think it is the best? Be sure to clearly explain your approach in language a non-statistician could understand.
