Revised research problem stated as a prediction study within a correlational design:
Predictor variables:
I think it's a natural desire to search for the cause to an effect. Understanding why and how things happen provides an explanation and the ability to somewhat control the future. If we can understand what caused an outcome then we have a greater chance of manipulating that variable in a similar situation to our benefit.
In research, however, the determination of a causal relationship as one of the primary research goals can limit the observations made or data collected (problems). If my research question involved only two variables - say, the effect of an inmate's exposure to natural light has on the number of violent incidences in which that inmate is involved per month - then I would end the research with data regarding quantities and types of natural light as well as reports of violent incident frequencies and types. If the research showed a positive correlation between natural light and lower violent incidents I could claim, based on this research alone, that there is a positive correlation. However, I think this would be an example of an accurate but invalid conclusion.
If the research topic has been previously explored and it has been determined there is a causal relationship then the new research can prove or disprove this. My research is the initial study and has a large number of variables. This seems fairly representative of social science research.
By using a correlational design model that is predictive I can include several variables that provide additional context to the overall research and inferences made. A regression model can communicate the role in which the predictor variables affected the outcome variable based on surveys, reports and interviews.
Questions:
Would an Ex-Post Facto, quasi-experimental design model, be appropriate given the number of "pre-existing" conditions participants may have?
Is a regression model the only communication tool, to express research outcomes, in prediction research as I've described above?
Predictor variables:
- exposure to natural light which includes an outdoor view including the horizon, expressed in hours/day
- reason for incarceration, expressed in type of non-violent offense
- length of time already served, expressed in years
Outcome variable:
- violent behavior with other inmates and prison officers, expressed in number of reported violent incidences per month
Population: Non-violent offenders in low-security prisons
Sampling Frame: 3 specific prisons, yet to be identified
Sample: sample of prisoners representing different races and ages
I think it's a natural desire to search for the cause to an effect. Understanding why and how things happen provides an explanation and the ability to somewhat control the future. If we can understand what caused an outcome then we have a greater chance of manipulating that variable in a similar situation to our benefit.
In research, however, the determination of a causal relationship as one of the primary research goals can limit the observations made or data collected (problems). If my research question involved only two variables - say, the effect of an inmate's exposure to natural light has on the number of violent incidences in which that inmate is involved per month - then I would end the research with data regarding quantities and types of natural light as well as reports of violent incident frequencies and types. If the research showed a positive correlation between natural light and lower violent incidents I could claim, based on this research alone, that there is a positive correlation. However, I think this would be an example of an accurate but invalid conclusion.
If the research topic has been previously explored and it has been determined there is a causal relationship then the new research can prove or disprove this. My research is the initial study and has a large number of variables. This seems fairly representative of social science research.
By using a correlational design model that is predictive I can include several variables that provide additional context to the overall research and inferences made. A regression model can communicate the role in which the predictor variables affected the outcome variable based on surveys, reports and interviews.
Questions:
Would an Ex-Post Facto, quasi-experimental design model, be appropriate given the number of "pre-existing" conditions participants may have?
Is a regression model the only communication tool, to express research outcomes, in prediction research as I've described above?
