Mirror Sites:
CSUDH - Habermas - UWP - Archives

California State University, Dominguez Hills
University of Wisconsin, Parkside
Soka University Japan - Transcend Art and Peace
Created: October 16, 2003
Latest Update: October 16, 2003
jeannecurran@habermas.org
takata@uwp.edu

Frequencies and Missing Values

Site Copyright: Jeanne Curran and Susan R. Takata and Individual Authors, October 2003.
"Fair use" encouraged.

This is a real important concept that most of us have straight, but I think we still need a little practice with it. Can't print yet from my machine, so you'll have to run these tables in the lab for yourself. I'll be a little late today, but Joyce promised to get you started.

The Issue:

Frequency tables give you percentages, and percentages are calculated by comparing the the number of respondents who chose a specific answer to the entire number of respondents who answerd the question. For example, if the question was gender, we would compare the number who answered female, to the entire number who answered the gender question.

A problem occurs in large surveys (whne some people are not asked some questions to try and keep the questionnaire from getting too long) and mistakes (like the one we made in leaving three questions out) when many respondents are simply not asked the question

In our data, questions 13, 14, 15, I think it was, were asked by only 41 of the questionnaires, even though 85 questionnaires were given out and responded to. For questions 13, 14, and 15, our results would be inaccurate if we compare the frequency of respondents who chose a specific answer to 85. We didn't have 85 answers for those questions. We only had 41. So we would have to compare the frequency for each chosen answer to 41. We would find that percentage in the Valid Percent column on output.

Practice:

Run analyses -> frequencies for questions 13, 14, and 15. On my program I have to pick analyses, link to frequencies, after I have chosen the variable. Then I just click on OK and the Output prints the percentages. The program automatically removes the missing values, and calculates the valid percent on the number of respondents who actually answered the question. For 13, 14, and 15 this will make a difference in the valid percent column.

Then write a paragraph, a brief paragraph, Denice, if you don't want a forty-page report, explaining what tables 13, 14, 15 tell you.

Then think about Mar Shay's question. Which answer will we have greater trust in, the one that calculates the percentages on 85 respondents, or the one that claucultates the percentages on 41 respondents? Why?

We'll review this in class this afternoon. jeanne