Tuesday, March 2, 2010

Content Analysis: Next Steps

You're proposing to do a content analysis. Dr. Hassett-Walker has approved (or approved with revisions) your study. Now what?

Here's what you do next (i.e., beginning your study):

(1) Do the search of your selected content. Ex: Searching in Westlaw for past articles in the NY Times and the Washington Post, during the years 2000-2010. Your search terms are "university" and "campus" and "violence" and "victims" (for a study - my hypothetical example - about the effect of violence on college and university campuses on primary and secondary victims).

Note: your search terms will be similar, probably, to some of your code list words.

Suppose 8,547 articles come up. You don't need that many. You need 100, but will randomly select 200 to account for some articles (let's say every other one) being non-relevant or less relevant.

(2) At Randomizer.org, generate a random number list. I just did this, and told Randomizer to give me 1 set of 200 randomly generated numbers, ranging from 1 through 8,547. Here's part of the list it gave me: 49, 2387, 4163, 3955, 4424, 2349, 77, 2422, 3480, 128, 5972, 1579, 3784, 2312, 624, 5602, 876, 2977, 2948, 6856, 2736, 7490, 2117, 7590, 4569, 2844, 2071, 5033, 6005, 2112, 4073, 3580, 7596, 7362, 7066, 1793, 5914, 2815, 7584, 3384, 118...

(3) Next, I find those articles that correspond to the numbers in the random number list. I copy/paste them into a Word or Word Perfect document.

(4) Now I begin reading the articles. Working in Word (rather than on paper), I use the Word artist function to color-code red all words and phrasings related to violence and shootings (e.g., shooting, gun, shots fired, etc.). I color-code green all words and phrasings related to victims (e.g., victim, victims, witness, students standing there watching, etc.) I color-code yellow all words and phrasings related to college or university campus (e.g., campus, college, university, classrooms, buildings, etc.). I color-code blue all words and phrasings related to programs the university or college put into place to help victims.

(5) I set up a database skeleton in SPSS. Before I can enter any data into the database, I must create my variables. These will correspond to my code list. EXAMPLE of variable name (variable #1) and corresponding label: ArticleNum, which newspaper did it comes from (1=NYT, 2=Washington Post). Next variable name: VictimNum, number of times victim or victims mentioned in article. Next variable name: VictimLang, language used to describe victims. Etc.

Note: any variables that involve counting will be numeric variables. Any variables that will have me inputting words will be string variables.

Remember: You set up your SPSS database in variable view.

(6) Once I have my SPSS skeleton set up, I can enter my data (words, language, counts of words) into SPSS. I can either do this once I'm done reading all the articles, or as I go. Remember: you enter your data into your database in data view.

(7) Once you've entered all your data, you run your preliminary and "main" analyses: frequencies, descriptives, and perhaps a correlation analysis. Review your results with Dr. Hassett-Walker. Do you think you are finding support for your research question or hypothesis?

(8) Once you have determined which relationships are significantly related to each other, go back to the color-coded newspaper articles that make up your data. Search for the text coded with the code words that pertain to those statistically significant relationships. That's the text you want to read and perhaps copy/paste into a separate word document. Boil it down into a succinct one to two-paragraph summary/summaries that further explains the statistically significant relationships you found with the quantitative data.

No comments:

Post a Comment