Crosstabulation

Overview Crosstabulation is treated under the following separate topics: Association Log-linear analysis Significance Survey research

Contents Frequently asked questions Bibliography

Frequently Asked Questions

• In crosstabs, should I ask for percentages by column, row, or total?
  It is customary for the independent variable to the the column variable, and for percentages to be in the direction of the causal variable. Thus in a table of gender by vote, gender would be the column variable and percentages would add to 100% in the columns. This is because you want to compare candidate percentages for men versus candidate percentages for women. Of course this all depends on the research purpose, and it is possible the researcher really does want row percentages in order to compare gender percentages across candidates. In this example, column percentages address the research question, "Are vote choices different for men compared to women?" Row percentages would address the question, "Is gender support different for Candidate A versus Candidate B versus Candidate C?"

  In general, given the research question, "Is the set of percentages of various categories of y the same or different within each category of x?", you want x to be the column variable and you want column percentages.

• How can I get table-related SPSS output by entering my frequencies rather than all my raw data?
  The researcher may need to do this to save time, or because one wants to compute association, significance, or some other parameter based on table frequencies without having access to the raw data. This is accomplished simply by weighting within SPSS. In the SPSS data editor, let the entries for data row 1 be 1, 1, n, to represent row 1, column 1, and the frequency (n) for that cell in your table. Let data row 2 be 1, 2, n for row 1, column 2, and the count for that cell in the table. Similarly fill in additional rows to correspond to each cell in your table. Then select Data, Weight Cases, and let Var0003 (the frequency variable in column 3, which you may want to relable as such) be the weighting variable. Then proceed to Analyze, Descriptive Statistics, Crosstabs to obtain output with measures of significance and association.

Bibliography

• Hellevik, Ottar (1988). Introduction to causal analysis : Exploring survey data by cross-tabulation. Second edition. NY: Oxford University Press. Contemporary social research series, 9.