Alexandra S.

asked • 08/09/19

Difficulty Formulating Hypothesis - Conflict Between Theoretical Model and Assignment Requirements


I am currently working on a project for my general linear models class.

The requirements of this section are as follows: "The problem should be stated in terms of causality where you are interested in the causal impact of ONE independent factor on ONE dependent variable, while keeping in mind your analysis will fall short of establishing causality." Additionally, the purpose of this project is to become familiar with multiple regression (OLS) and testing its assumptions to determine if it is a good model to use for the research question.

I already know that OLS regression is not the appropriate model to test what I am interested in, this is irrelevant as this project is for learning purposes.

Essentially, what I am trying to test is a theory that states that familial power relations (i.e. occupational status/prestige of the father vs. that of the mother, resulting in "balanced" and "unbalanced" families) when an individual is growing up is related to the gender distribution of perceived risk of victimization in adulthood.

Previous literature has repeatedly found that there is a gender gap between males and females average perceived risk of victimization, with females reporting significantly more.

Therefore, what I (if I were allowed to start with the model I would like to use) ideally am hypothesizing is "The relationship between gender and perceived risk of victimization should be reduced in 'balanced' family structures."

However, this is where I am struggling. I need to reformulate this idea into a hypothesis testing ONE independent variable (gender OR family structure) on the dependent variable (perceived risk of victimization).

I am just not sure how to do this, and nothing I have thought of seems sufficient. I know I am going to have to control for one of the two IV's, but is there a way to reflect this idea that I am interested in in a hypothesis that fits the assignments requirements?

PLEASE help :( & Thank you very much for your consideration!

1 Expert Answer


Thomas C. answered • 08/10/19

New to Wyzant

Clear Communicator for Math, Science and Writing

Alexandra S.

Hello Thomas! Thank you for your thoughtful response. In terms of operationalization, I am utilizing the 2018 General Social Survey, which includes a questions regarding perceived risk of victimization (i.e. Are you afraid to walk alone at night -- Within a mile radius of your home?) I am aware this is limited, however fear of crime/perceived risk of victimization is very difficult to measure, and prior criminological studies have utilized similar questions (hence why I am calling it perceived risk of victimization rather than fear of crime, because some have argued questions such as the one I am utilizing do not accurately measure "fear of crime," per se, but instead measure an individuals perceived risk of victimization in a given situation). In terms of measuring familial class relations, there is a question asking whether the mother/father was self-employed or had a supervisor, and for the mother there's an additional variable that accounts for whether she was unemployed or didn't work for a period of a year or greater when the respondent was 16. Together from this, I can recode the father and mother (separately) into the obey or command classes (not relevant to my question at all) and term the family balanced or unbalanced based off of that. For this project, the operationalization and/or measurement of our variables isn't necessarily important (we're not graded on that aspect). The most important part is that we are running OLS regression on one independent variable on one dependent variable, and test the assumptions and correct the issues if violated. My research question essentially is "Why are women, on average, perceive significantly more risk of victimization than do men?" In this case, my ideal hypothesis that I would like to test, as previously mentioned, is "the relationship between gender and perceived risk of victimization should be reduced in 'balanced' family structures." I simply need help trying to figure out how to "dumb it down" to be one independent variable and one dependent variable so that I can do the project. Doesn't matter if it's correct, because that's the point of the project: to test the assumptions and see if they hold, and make recommendations/adjustments accordingly. Thanks!


Alexandra S.

Also, I'm not sure why this posted as one large paragraph. I apologize for the inconvenience.


Thomas C.

From my reading of how you're deriving your independent variable, I think the basic problem is you're trying to encode a binary variable. You *can* peform linear regression for binary variables, but it's not particularly appropriate most single-variable OLS setups. That and your question is way way WAY WAAAAAAY more complicated than one independent variable and one dependent variable. Maybe your encoding thing will yield some interesting results, but who knows? That said, in order to accomplish what you want, you have two basic choices. 1) Find two data columns/questions in the GSS that get as close as possible to what you're asking. 2) Develop an index - a single variable that is a function of some set of other variables - for your independent and dependent variables. The S&P 500 is an example of an index. -Thomas


Still looking for help? Get the right answer, fast.

Ask a question for free

Get a free answer to a quick problem.
Most questions answered within 4 hours.


Find an Online Tutor Now

Choose an expert and meet online. No packages or subscriptions, pay only for the time you need.