The Quantitative Methods Paper is made up of five sections:
Introduction
Lit Review
Methods
Analysis
Discussion and Conclusion
These sections are discussed below.
Sure the introduction to any paper introduces your paper to the reader, but the introduction section is more important than that to an academic paper (yes, that's what you are writing). There are many papers and journals out there in the world for social scientists to read. Your introduction needs to convince the sociologist that he or she needs to spend precious time reading YOUR paper. If you can't show why studying your dependent variable is important in a couple of paragraphs, then you need to get a new dependent variable. Why are things interesting or important? Perhaps it is because the topic is controversial (Some people believe/feel/act one way, and some another.). In any event that is the point of the intro.
In this section, the main question that needs to be answered is what has been written before on your topic? In particular, you are interested in what has been written concerning any relationship between your dependent variable and your independent variables. In a normal academic paper, you need to demonstrate that you know every detail of the material important to your hypotheses. However, in this class I am only asking you to produce a minimal literature review.
This is an extremely unlikely occurrence. I would begin by looking for articles using alternate terms which have the same meaning as your concept. I would also talk to the professor. He is wise in the ways of science and can probably help.
At the end of the lit review, you state your hypotheses.
The method section has three parts:
Example:
This analysis utilizes interview data collected by the National Opinion Research Center (NORC) in the 1994 General Social Survey (hereafter GS S). The GSS, a nationwide annual survey, offers the advantage of multi-stage probability sampling and can be considered representative of English-speaking, noninstitutionalized adults (18 years of age and older) living in U.S. households. (For more detailed information on the GSS, see Babbie and Halley [1994].) This examination of the relationships between x, y, and z relies on a subset of 958 of the 2992 original respondents. The data extract includes only questions asked on both interview ballots B and C for Version 2 of the 1994 GSS. This provides the researcher with a continuous set of questions with a lower number of missing cases; however, the trade-off is the lower number of total cases. Following is a brief description of the variables considered and of the frequency distributions for these variables.
In this class we are going to concentrate on making sure you can calculate univariate frequency distributions, crosstabular analysis, including control variables, and regression analysis.
The analysis section starts off with you restating your hypotheses. Then you begin your examination of whether those hypotheses were supported by the data.
Using the output from the 1st crosstab, tell the reader if it was supported.
Then you show the reader how you know it was supported (Hint: Talk about the %s in the crosstab table)
Then you tell the reader if the results you see are statistically significant.
Display the Chi-Square value (for this course use the Pearson's Chi-Square) and the p-value (normally the Asymp Sig; however, if you have 2X2 table use the Exact Sig (1-sided) result).
If your table has significant results talk about the strength of the relationship.
Do the same thing as crosstab 1.
The controlled crosstabular analysis is also referred to by the phrase "the elaboration method". While we will have gone over this in class, you may want to look that phrase up in a couple of methods texts for a more in depth discussion.
The first thing you have to do is choose which of the two hypotheses you tested is your primary hypothesis (HINT: it is most likely the hypothesis tested in crosstab 1.
You are then going to control the relationship between the variables in your primary hypothesis by looking at the relationship between your independent variable and your dependent variable at every level of your control variable. What this means is that the computer builds a crosstab table to examine the relationship between your IV and DB for each responce category of the control variable. For example, if I were interested in the relationship between political party (PARTYID) and frequency of sexual relations (SEXFREQ) and I controlled that relationship by sex. SPSS would build a table crossing PARTYID and SEXFREQ for males and another table crossing PARTYID and SEXFREQ for females. If I had controlled by AGE instead, SPSS would have built a table crossing PARTYID and SEXFREQ for each age category. Each of these separate tables will have its own chi-square statistics and its own lambda and/or gamma statistics (if you asked SPSS to calculate statistics).
Now, for the write up there are just about 5 different variations for the controlled crosstab write-up. You will need to see which one fits your situation. One of the major factors in deciding which variation you use will be the relationship you originally observed between your IV and DV in your earlier crosstabular analysis. Here we go:
The first two cases occur when your initial crosstabular analysis weren't significant.
If the original crosstabular analysis relating your independent variable and dependent variable WAS NOT SIGNIFICANT and you look at each crosstab table for every level of your control variable and they are still not significant, you can then say: "My original relationship was not significant and when controlled by my control variable, Z, the relationship remained non-significant.".
If the original crosstabular analysis relating your independent variable and dependent variable WAS NOT SIGNIFICANT and you look at each crosstab table for every level of your control variable and one or more of the tables IS SIGNIFICANT, then you can say: "My original relationship was not significant; however, controlling by Z revealed a suppressed relationship between X and Y".
The next three cases occur when your initial crosstabular relationship was significant.
If the original crosstabular analysis relating your independent variable to your dependent variable WAS SIGNIFICANT and you look at each crosstab table for every level of your control variable and ALL of the tables STILL SHOW A SIGNIFICANT RELATIONSHIP, then you can say: "My original relationship was significant and when controlled by Z remains significant. The relationship between X and Y is not caused by the influence of Z".
If the original crosstabular analysis relating your independent variable and dependent variable WAS SIGNIFICANT and you look at each crosstab table for every level of your control variable and ALL of the crosstab tables ARE NOT SIGNIFICANT, then you can say: "My original relationship was significant, but controlling for Z, the relationship now appears to be spurious. Z appears to be responsible for the observed relationships between X and Y."
Lastly, we have the tricky one--the mixed case. This case is, of course, what most of you are likely to see when you look at your controlled crosstabular analysis. IF the original crosstab comparing your independent variable and dependent variable WAS SIGNIFICANT and you look at each crosstab table for every level of your control variable and see that SOME of the tables ARE SIGNIFICANT and SOME ARE NOT SIGNIFICANT, then you will need to make a judgment call. Here's the judgment:
Were there enough respondents in each of the controlled crosstab tables?
WHY IS THIS THE IMPORTANT JUDGMENT CALL? We know that as your N in a crosstab table increases that smaller differences are more likely to be considered statistically significant. It is possible that your data still exhibits the same patterns (in the percentages) that you saw in your earlier crosstab , but since your sample is divided across several tables it won't be statistically significant.
IF you believe that the table does show the same pattern, but fails to be significant due to a small number of respondents. You may argue that. If you can argue that for all the controlled crosstab tables that aren't significant (if there aren't too many), then you could state that "It appears that the relationship between X and Y persists when one looks at the patterns in the column percentages; however, some of the controlled crosstab tables are not statistically significant. Still, I would argue against calling this a spurious relationship. My reading is that the relationship between X and Y is not truly caused by Z."
OTHERWISE, you will need to argue that the control variable mediates the relationship. That is, the control variable really helps delineate in which situations the relationship holds. For instance, you might find that your relationship between X and Y holds for whites but not for blacks or holds for males but not for females. This can be very important information. In this case you will need to report the significant relationships like you did in Crosstab 1.
We didn't get to regression this year, however, I would like to point out a few things that you will have to interpret.
We look at the F statistic (and its significance) to determine if the model is significant.
We look at the r-square to determine the amount of variation in the dependent variable that can be explained by the variables in the model.
We look at the t-statistic (and its significance) for each independent variable. These tell us whether each IV is significantly related to the DV, controlling for the other variables in the model.
We look at the b line to figure out the slope of the line.
We look at the Betas to determine which variable has the most strongest relationship with the dependent variable.
As opposed to the rest of the paper which tends to be heavily formatted, the conclusion section is yours to say what you want. HOWEVER, you must say something.
Traditionally, the conclusion section begins one more time with a statement of your hypotheses. This is followed by a summary of your findings. Were your hypotheses supported or not?
The conclusion is more than just a summary, however, because you also get to speculate on how to do things better. For instance, it could be the case that your hypotheses weren't supported, but you really believe that the relationship exists. You could then bring up issues of validity and reliability. You could state that future research should ask more or different questions. You could state that future research should use more variables or add different variables. You could argue that the sample was poor. It's your opportunity to brainstorm on how future research should be done.
Why do this? Well, the idea is that we, as social scientists, stand on the shoulders of the others that have come before. We owe future researchers who are reading your article to glean some knowledge about how to approach your concept at least some guideposts as to what we think worked, what didn't work, why, and what we would do if we were going to continue to do research on your topic.