Hello guys,
I posted something in reference to what I am going to ask here in a past post but I think it was necessary to create a new post. I want to start by saying that this project it’s very complicated and that this post relates more to the designing part of project. Any idea/suggestion will help, any!
I have a dataset containing Use of Force (UOF) incidents in a mental health facility. When the mental health staff have to use force against a patient, that incident is recorded. Details about the staff (can be more than one staff), patient (can be more than one patient) involved, as well as “environmental” characteristics of the incident (for example: location-housing unit, shift, population) are recorded. Below is a sample of how the data looks for an incident (not all variables are included).
The ambitious general goal of this project is to develop a model that will predict which mental health providers are more like to use force. My first approach was to collapse the data base on the officer, and the inmate level creating a variable of the count of UOF for that year. For example, when I collapse the data to the officer level, each row represents an officer. I did the same thing for the inmate. Using these datasets, I just run OLD regression to see potential predictors of UOF variable (the number for the year DV). For example, I found that staff with a higher tittles have higher numbers of UOF in the year. Likewise, more experience staff (years in the job) have fewer UOF. Some predictors for patients are numbers of admission in the past, etc. I also collapse the data to the “incident” level just to run frequency tables at some crosstabs. Each row is an UOF incident. For example, I know that more UOF incidents occur in the morning shift compared to night (since all patients are sleeping)
Here are my questions: how can I combine the effect of both: officer and inmate, and even the “environmental” characteristics of the incident? What statistical model should I use? Should I use a mixed effect model? How should the data be structured to conduct analysis? If this data does is not sufficient? What type of data should I collect? I have some ideas on what to do but I would prefer to hear from you guys first.
Best,
Marvin
I posted something in reference to what I am going to ask here in a past post but I think it was necessary to create a new post. I want to start by saying that this project it’s very complicated and that this post relates more to the designing part of project. Any idea/suggestion will help, any!
I have a dataset containing Use of Force (UOF) incidents in a mental health facility. When the mental health staff have to use force against a patient, that incident is recorded. Details about the staff (can be more than one staff), patient (can be more than one patient) involved, as well as “environmental” characteristics of the incident (for example: location-housing unit, shift, population) are recorded. Below is a sample of how the data looks for an incident (not all variables are included).
| FORCE_ID | UOF_DATE | SHIFT | FACILITY | PERSON_TYPE | STAFF_ID | STAFF_TITLE | STAFF_EXP | STAFF_DOB | OFFICER_GENDER | PATIENT_NAME | PATIENT_DOB | PATIENT_RACE | PATIENT_SCORE | PREVIOUS_UOF |
| 10045 | ########## | 3 TO 11 | MMD | EMPLOYEE | 1025 | DOCTOR | 4956 | 11/14/1977 | F | |||||
| 10045 | ########## | 3 TO 11 | MMD | EMPLOYEE | 4584 | ASSISTANT | 5214 | 1/8/1985 | F | |||||
| 10045 | ########## | 3 TO 11 | MMD | INMATE | MARIA | 6/16/1985 | BLACK | 5 | 15 | |||||
| 10045 | ########## | 3 TO 11 | MMD | INMATE | JENNIFER | 5/10/1999 | WHITE | 20 | 1 |
The ambitious general goal of this project is to develop a model that will predict which mental health providers are more like to use force. My first approach was to collapse the data base on the officer, and the inmate level creating a variable of the count of UOF for that year. For example, when I collapse the data to the officer level, each row represents an officer. I did the same thing for the inmate. Using these datasets, I just run OLD regression to see potential predictors of UOF variable (the number for the year DV). For example, I found that staff with a higher tittles have higher numbers of UOF in the year. Likewise, more experience staff (years in the job) have fewer UOF. Some predictors for patients are numbers of admission in the past, etc. I also collapse the data to the “incident” level just to run frequency tables at some crosstabs. Each row is an UOF incident. For example, I know that more UOF incidents occur in the morning shift compared to night (since all patients are sleeping)
Here are my questions: how can I combine the effect of both: officer and inmate, and even the “environmental” characteristics of the incident? What statistical model should I use? Should I use a mixed effect model? How should the data be structured to conduct analysis? If this data does is not sufficient? What type of data should I collect? I have some ideas on what to do but I would prefer to hear from you guys first.
Best,
Marvin