«Social Psychology (Soc 220) J. Lashbrook Data Analysis Excercise Spring 2004 Social Structure—Personality: What is the relationship between social ...»
Social Psychology (Soc 220) J. Lashbrook
Data Analysis Excercise Spring 2004
Social Structure—Personality: What is the relationship
between social class and child-rearing values?
The sociologist, Melvin Kohn, argued that people’s locations in social structures,
particularly the occupational structure, influenced the values they would stress for their children because variations in structural locations exposed them to different experiences.
His research, however, was done some time ago. Much has changed in our social world since then. Does the original relationship still hold?
OBJECTIVESThis assignment is asking you to generate statistical results, analyze and interpret the data produced, and then use your results to take a brief quiz on Angel, the exact directions for which will be following shortly. Analytically, we will answer
the following 3 questions:
• How important to people is stressing to their children that they think for themselves?
• Does social class influence this child-rearing value of thinking for yourself?
• Has this influence changed over time?
Other learning outcomes for this assignment include honing analytical skills associated with generating, reading, and interpreting rudimentary forms of data analysis.
A PRIMER ON THE LOGIC OF SOME BASIC FORMS OF DATA ANALYSISIntroduction Most science is predicated on the belief that the world is not just a random, chaotic mess, but rather things happen in certain ways and for certain reasons. Therefore, scientific research is often categorized by two major objectives—to describe and to analyze. Scientists will document the order or patterns in the world (our descriptive objective) and the reasons behind such patterns (our analytical objective). Social scientists are no different. We’re interested in whether or not there are patterns or relationships among the social phenomena on which we collect information. Analyzing this information, whether quantitative or qualitative in form, involves a conversation between you and the data. As researchers, we pose questions guided by our conceptual understandings of the subject matter and then use various statistical strategies to see what the data suggest in the form of answers. We have a variety of tools for collecting information or data; surveys are one major conduit for such information. Survey research translates our phenomena of interest into variables. A variable is anything that can vary, i.e., take different values, and it represents our way of measuring the concepts in which we are interested. For example, gender is a variable because people can be male or female. Information on variables like gender from surveys can be entered into a computer program by assigning numerical values to people’s answers. Then the data can be analyzed in the search for patterns.
Univariate Analysis Now, to look for patterns, I might examine the variation on just this one variable by looking at the distribution of cases [respondents] across the variable’s response set (we call this doing “univariate analysis” by using a statistical program to generate a “frequency distribution”). In reference to gender, doing that would tell me how many (or what proportion in %) of my survey respondents classified themselves as male or female. Thus, I get a portrait of how gender is distributed in my sample. While it doesn’t take me real far, this type of analysis is still useful in its own right.
Since we’re interested in beliefs, behaviors, and feelings, here’s another real example from some actual data taken from the same survey you’ll be analyzing. It’s called the General Social Survey (GSS). Further below you’ll find a brief description of the GSS. The table below presents the distribution of people’s answers to the question, ‘in general, do you find life exciting, pretty routine, or dull?’ Table 1: Respondents’ Self-Reported Level of Excitement in Life—1998.
Trend Analysis Another useful way to look at frequency distributions is to examine how they change over time. We call this trend analysis. For instance, keeping with my above example, if I had surveys from different years that asked the same question on how exciting people find their lives, I could document whether or not there’s been any change.
In fact, over the years, the GSS has asked people this question. Here’s an example of looking at some trend data on life excitement using their results (Note: GSS aggregated
their results in the first time-period):
Table 2: Trends in How Exciting Respondents Find Their Lives: 1988-2000.
Interpreting Data Statistical software generates the results like those shown above. However, I’ve taken those results and dressed them up a little in a couple professional-looking tables (when you do your actual analysis, you’ll see how the output is different from these tables). But while the computer does the grunt work, you must still interpret what the numbers are saying. Being able to clearly and concisely interpret results for an audience is an important skill to develop for many occupations. For these simple tables, probably 2-4 sentences would suffice. You won’t see this in a textbook, but here are a couple guidelines for interpreting this kind of information. First, always provide the audience with some of the contextual detail (i.e., what is the data, where does it come from, when, etc.). Second, construct your interpretation from the “general to the specific.” In other words, start out with a general statement and then move to including some specific figures. Interpreting this kind of frequency distribution often makes use of a particular statistic called the mode. A mode “is the most frequent, most typical, or most common value in a distribution” (Levin & Fox, 1994, p. 77). In other words, for this data, it’s the category with the largest percentage of respondents.
Using these guidelines, interpreting Table 1 might go something like this:
Respondents from the 1998 General Social Survey are nearly split on how exciting they find life. A slight majority of GSS respondents, based on their self-reports, find life either routine (49.4%) or dull (5.5%). On the other hand, nearly half (45.1%) report that they find life “exciting.”
Here’s what we might say in looking at the trends in Table 2:
The percentage of GSS respondents that report life is “exciting” increased from 1988-2000. While less than half of those surveyed in the 1988-1991 period (44.7%) reported life is “exciting,” slightly over half (52.1%) were excited in 2000. The greatest increase happened between 1998 and 2000. Overall, those who found life either “routine” or “dull” declined in this same time period.
BIVARIATE ANALYSIS: LOOKING FOR RELATIONSHIPS BETWEEN TWO VARIABLES
That’s a brief introduction to some basic types of data analysis, but if we want to see how variables are related to each other, we can use a technique called cross-tabulation. A "cross-tab" is a table that presents the distribution (in frequencies and/or percents) of one variable across the categories of another variable(s) (e.g., what percentage of men find life exciting compared to women?). Since it let’s us look at two variables, we label this a bivariate analysis.
Typically, in crosstabs and many other statistical techniques, we conceptualize the relationship between the two variables in terms of one influencing the other. The language we use to capture such relationships is to call one variable an independent variable (IV--it's doing the influencing) and the second variable a dependent variable (DV--it's the one that is being influenced). To run a crosstab you decide on two variables that you think might be related to one another. Then, drawing from a conceptual framework, one next states a hypothesis for the relationship between your chosen variables. A hypothesis is an educated guess about what you think you will find. Hypotheses should always state a specific relationship and specify the comparison (e.g., Women are more likely than men to find life exciting.).
In order to understand the logic of crosstabulation, think of a psychologist’s experiment. Say she wanted to examine how a particular studying strategy affects students’ test performance. The previous sentence casts study strategy as my IV (it’s doing the influencing) and test scores as the DV (it’s being influenced). She might hypothesize that students using the special study strategy are more likely to do better on the test compared to those students who didn’t use the strategy. So she sets up a situation where one group is given the special study strategy and the other group is not. She then tests the students. How does she figure out whether the study strategy makes any difference? She needs to compare the two groups, of course. For example, she might calculate the mean test score for each group to see if one is higher than the other. Or she might divide up test scores into “low,” “medium,” and “high” categories and see what percentage of students falls into each one for the two groups. This latter strategy is what crosstabulation analysis is all about.
For this assignment, we’re interested in assessing whether social class makes a difference in childrearing values. Kohn originally stressed occupational location, but for this assignment we’ll use another rough structural proxy—social class location. While it’s not perfect, it’s reasonable to think that people in higher social classes are more likely to be in the kinds of jobs that stress autonomy and self-direction. If that’s the case, who will be more likely to stress “thinking for him/herself” as an important value for kids to develop? Stating an answer to this question is formulating a hypothesis. But you need to be clear on thinking about what is the independent variable and the dependent variable (hint: is social class doing the influencing or is it the thing being influenced?).
To show you what a crosstab looks like and how it’s interpreted, I’ll use another topic related to childrearing and which also draws on the social structure—personality approach. Sociologists know that social class is connected to many things. An analytical question might be, does social class make a difference in how parents discipline their children? Drawing from Kohn, middle class parents, because they tend to work in less regimented environments, might actually be less regimented in their approach to discipline. A concrete variable could be whether or not someone favors physical forms of punishment.
Let’s say that the actual measures for these two concepts are educational level as a proxy for social class and the degree to which a parent believes in spanking as a form of discipline (measured with a response set of “strongly agree,” “agree,” etc.). (Before going any further, let me stress that these measures are not “perfect” indications of the concepts we’re interested in. There is no such thing as a “perfect” indicator.
But they are decent approximations and social scientists work with these types of indicators all the time.) I hypothesize that parents with higher levels of education will be LESS likely to favor spanking as a disciplinary technique compared to parents with a lower level of education. Here, education is the IV and attitude towards spanking is the DV. Below are results generated from the GSS for their 1993 survey.
Table 3: The Relationship Between Educational Level and Attitudes Toward Spanking for Discipline—1993 GSS
Interpreting the results. Here again, the computer does the number-crunching for you, but you still have to interpret what the tables are “saying.” Remember what you’re after: you want to see if the IV (education) makes any difference in the DV (spanking attitudes). To do that, you want to compare percentage differences between the categories of the IV (e.g., compare those with lower levels of education against those with higher). [Note: to do this properly, you want to make sure each category of the IV adds up to 100%. Analytically, you’re asking what % of people with a high school education or less favor spanking versus what % of people with a college education?] In this example, if you compare DV categories, you see that a higher percentage of people approve of spanking than don’t. I’m not saying that’s not interesting, but it doesn’t address the conceptual question we began with: does social class make a difference in childrearing attitudes? To answer that, we look at differences between educational categories. Besides, you could have already noted the differences in those who approve or disapprove spanking in a univariate analysis (frequency distribution).
There are three basic questions we can answer from this type of crosstabs output. You should frame
your interpretations around these 3 questions:
Is there a relationship between the IV and DV? (You’re looking to see if there is enough of a difference in percentages between the columns to matter. Some guidelines are found in the third bullet here.) What is the relationship? (Here you want to put in sentence form what it is that the table is telling you.
Be careful how you state it because you may end up saying something that the results don’t show. You always state the % of people in the IV [from below, “…two-thirds (66.3%) of those with a 4-year college degree or higher…”]. Also, please be aware that I’ve given a simple example here. Many of our variables of interest have more than two categories. If your IV has 3 or more categories, the logic is still the same. You want to just compare across the IV categories. If the categories on each end seem to represent opposite ends of whatever variable you’re looking at, then you can look at those.