The first thing we need to cover are the three main types of variables. The first type are independent variables (called IV's for short). These are predictor variables, or those that we expect to affect other variables.
The second type are dependent variables (called DV's for short). These are outcome variables, or those that we expect to be affected by other variables.
IV's and DV's can be summarized pictorially as: IV ---> DV
To help you, they can also be imagined as: cause ---> effect
The third type of variables are mediating variables. Mediating variables are those that we expect to affect some variables and to be affected by other variables. They can be summarized pictorially as:
IV ----> mediating variable ----> DV
Here's two examples to clarify the types of variables:
If I think that girls are more likely than boys to become secretaries, the IV is gender and the DV is type of career, as below:
Gender ----> Type of Career
If I feel that girls are more likely to be socialized into "feminine" careers (while boys are socialized into more "masculine" careers) and that makes them more likely to become secretaries, the IV is gender, the mediating variable is type of socialization, and the DV is type of career.
Gender ----> Level of Socialization ---> Type of Career
Recap Q1: I hypothesize that smoking causes cancer. Which of the following variable types does "smoking behavior" fit under?
A. IV
B. DV
Now, you may have noticed that while I felt that girls would become secretaries, my variables weren't "girl" and "secretary." Instead, they were gender and career type. This is because "girl" isn't a variable, it's an attribute. Variables must be able to vary, hence their name. Variables are made up of attributes, so that a given child's gender could be either the attribute "girl" or the attribute "boy." (We're not going to get consider cases where someone could be identified as both or neither gender).
Likewise, the variable career type could be secretary, truck driver, doctor, dentist, construction worker, or a host of other attributes. When we want to talk about all those occupations using one word, we call them career type.
Recap Q2: Which of the following is NOT a variable?
B. level of satisfaction with supervisor
D. number of years at the company
Now, on to the final part, relationships. When we say we're "in a relationship with someone," we're saying that we are somehow associated with that person. Sometimes, others can tell something about that person based on the fact that they are in a relationship with us; it would be hard, for example, to imagine a staunch conservative marrying a devoted liberal. Knowing that a person is a staunch conservative helps us picture their mate.
It works sort of the same way with variables. In research, two variables are related if knowing the value (or attribute) assigned to one helps us predict the value of the other. Assume for a moment those who studied at least five hours for the midterm exam in this class tended to get better grades, as in the table below:
IV: number of hours studied for the midterm exam < 5 5 or more DV: grade C or less 9 1 B or better 1 9
Now, what grade would you predict that your classmate Joe received? The first thing you would want to know is how many hours Joe studied. If you learn that Joe studied for seven hours, you would guess he received a 'B' or better. The reason for this choice would be that 9 of the 10 people who studied for five or more hours got a 'B' or better. That's what bettors call darn good odds. Even if Joe is the one person who got a 'C' or less, you had a 90% chance of guessing his grade based on knowing just one factor about Joe (the number of hours he studied). We could easily say that number of hours studied and midterm grade are related. Of course, most relationships in the social sciences aren't quite so strong, which makes accurate predictions a bit harder to do. When knowing the value of one variable doesn't help us predict the other at all, we say the variables aren't related. To illustrate this, try to predict the grade received by a male classmate, using the table below.
IV: gender of student male female DV:grade C or less 5 5 B or better 5 5
There is no relationship here. Gender does not appear to be related to midterm grade; males and females are equally likely to get a 'B' or better.
One way to look at relationships is to consider it as a betting game. While you might be willing to put money on a bet regarding a student who studied for seven hours, you would probably not want to put a whole lot of money on a student about whom all you knew was his/her gender. It would be too risky of a bet. Knowing gender doesn't help you predict grade enough to bet any money on the outcome. Whenever you would be willing to bet money on the outcome, the variables are related.
Recap Q3: Which of the following scenarios best illustrates two variables that are related?
A. gender and major, where 50% of sociology majors are male
B. GPA and career choice, where 1/2 of passing students become program administrators
C. major and satisfaction with career, where 75% of sociology majors are satisfied with their career
D. marital status and decision to go to college, where half of married people go to college
You have now completed the section on "Variables and relationships." There will be more on relationships in the statistics section. Click here to return to the main menu
Page last updated: April 15, 2008