CROSS TABULATION --create tables to describe the relationship between two variables.

1.  When you select the CROSS TABULATION option, a window will appear for entering the row variable and the column variable.  Make your dependent variable your row variable; make your independent variable your column variable.
a.  Note that you can enter subset variables in the CROSS TABULATION option just as you did in the UNIVARIATE STATISTICS option.  This is one way to eliminate Adon=t know@ and Ano answer@ responses.
b.  In addition to subset variables, this window allows you to indicate control variables. If you select a control variable you will look at the relationship between your independent and dependent variables when some additional variable(s) is controlled.  For example if the independent variable is R.HME WRK How much of the housework does the respondent do? and the dependent variable is HAP.MARR.? How happy is your marriage?, a control variable might be SEX.  MicroCase would then produce two tables, one for males and one for females.
c.  When you subset a variable, you restrict the analysis to a part of the sample. When you control a variable, you analyze all of the sample, but one part at a time.

2.  When you have entered your dependent (row) variable, your independent (column) variable, any subset variables and any control variables, click on OK in the upper right hand corner of the window.  MicroCase will take a few seconds, but a frequency table will appear.
a.  Convert the frequency table to a percentage table by clicking on COL% in the menu on the left side of the table.
b.  Always percentage within the categories of your independent variable.  If you always make your independent variable the column variable and always choose COL% you will have an easier time reading tables.  If for some reason, however, you have made your independent variable the row variable, then you would choose the ROW% option.
c.  Do not interpret the frequency table; always convert the frequency table to a percentage table.

3.  If you haven=t already done so, eliminate Adon=t know@ or Ano answer@ categories now. Using your mouse, click on the response(s) you want to eliminate.  The column or the row should be highlighted in color.  Then click on the COLLAPSE button on the left side of the screen.  A small window will appear which will ask whether you want to combine categories or whether you want to convert a category to missing data.  To eliminate Adon=t know@ or Ano answer@ select the CONVERT TO MISSING DATA option.  Instantaneously, the row (or column) disappears from the table and the percentages are recalculated.

4.  Percentages are very unreliable when the sample sizes are small.  Many scholars would argue that you should have close to 30 people in each category of the independent variable; this is a very strict rule.  An easier and more common version of the rule is that you must have at least 10 people in each category of the independent variable.   Especially when you are controlling one or more additional variables, it is common to have categories of the independent variable with less than 10 people in them.  In such instances consider whether you can combine independent variable categories to make a new category.  It probably would not make any sense to combine Jews and Catholics to make a larger religious preference category; Jews and Catholics are too different.  It might make a lot of sense to combine Strongly Agree and Agree into an Agree category, however.  To combine two columns or rows, highlight them by clicking on their label; then click on the COLLAPSE button on the left.  In this case, you=ll select the COMBINE CATEGORIES option.
a.  Note that MicroCase has given the new category one of the old labels.  You should give it a new label; to do so, click on the MicroCase suggestion and use your backspace or delete key to remove the suggested label; then type in your own label.  (MicroCase labels are capitalized; you might make your labels lower case to remind you that you=ve collapsed categories.)

5.  If you have selected a control variable, you move back and forth between the control tables by clicking on the left and right arrows on the left side of the screen.

6.  If you want a visual depiction of the table, click on ASTACKED BAR GRAPH@ from the options listed on the left side of the screen.

7.  The gray menu on the left side of the screen provides an option for EXPECTED FREQUENCIES (the frequency table is recalculated to show what the frequencies would be if there were NO association between the variables; this is a necessary step in calculating chi-square).

8.  The side menu also provides a STATISTICS option which will provide you with the calculated chi-square value (with degrees of freedom and the probability that a chi-square value of that magnitude could be due to random sampling error alone).  In addition to chi-square the STATISTICS option provides some of the nominal level measures of association (Lambda, Cramer=s V, Contingency Coefficient) and some of the ordinal level measures of association (Gamma, tau-b, tau-c, Somers= d).
a.  The ordinal level measures of association include probabilities that a coefficient of that magnitude could occur by random sampling error alone
b.  Chi-square probabilities can be used to evaluate whether Cramer=s V and the Contingency Coefficient are statistically significant.
c.  There is no sampling distribution for lambda, so you cannot assess the probability that a given lambda coefficient is due to random sampling error alone.
d.  If your analysis includes a control variable, remember that the Statistics are for the partial table only.  MicroCase calculates a partial gamma and gives the overall gamma so you can compare them.