Cross Tabs

1. Always click STATS and checkbox on chi-squared and tau-c when you do cross tabs and include these below the crosstab table, or else write the correlation and significance (and the chi-squared significance if there is a zero linear correlation) next to a cross tab, and if variable numbers are not printed on the cross tab, make sure to write them in.

2. When you get your output, click "Case Processing Summary" and delete with control-x (Edit/Cut); also Click "Crosstabs" and delete. THESE JUST CLUTTER YOUR OUTPUT AND I DO NOT WANT TO SEE THEM IN YOUR PAPERS. The validity rate reported under this summary is merely the percentage of case coded. Pay no further attention to it!

2. Percentaging is not required, but may be useful. If you percentage by rows, you can compare increasing or decreasing percentages in the columns. If you percentage by columns, you can compare increasing or decreasing percentages in the rows. NEVER PERCENTAGE BY TOTAL or BY BOTH ROWS AND COLUMNS: make up your mind which makes more sense, or percentages by the variables that has more categories.

Bar Graphs - the essentials

1. When presenting a crosstab plus statistics, if it makes sense you may substitute a bar graph for the crosstab (but keeping the statistics output with the graph) To do so in spss click Graphs/Bar/Clustered/Define/ and either N or % of cases and click one variable into "Category Axis," the other into "Define clusters by" BUT:

2. ALWAYS click Options at this point and unclick "Display groups defined by missing values"

3. Remember to ALSO RUN crosstabs for this pair of variables and include the statistics boxes with the graph.

Correlations

1. For Correlations, group the variables into sets. Make one big correlation table in Spss: Statistics/Correlate/Bivariate, then CLICK ON EACH OF YOUR VARIABLES IN ORDER in set 1, set 2, etcetera, according to topics. The correlations WITHIN SETS should give you information about reliabilities (or how variables with the same topic measure different aspects of the topic). The correlations BETWEEN SETS are usually the ones that help you TEST YOUR HYPOTHESES.

2. When you get your output, click "Case Processing Summary" and delete with control-x (Edit/Cut); also Click "Correlations" and delete. THESE JUST CLUTTER YOUR OUTPUT AND I DO NOT WANT TO SEE THEM IN YOUR PAPERS.

3. Typically you only need one correlation table (prefereably tau-c) in your paper and you might choose to skip the correlation table altogether. Alternatively, you might have one such table to show the reliability of each set of variables. Or you might present a correlation table for all of the variables with significant LINEAR relationships.

4. If variable numbers are not printed on your correlation table, make sure you write them in.

5. Beware variables with small sample sizes! Note how they lower the significance of correlations: and their correlations suffer random fluctuations. If you are using variables in the range 481-528, for example, look at the codebook and use only the average scores for boys and girls, which have overwhelmingly larger sample sizes.

Printing

1. To print a large table, click the table, then Find/Print/Properties and click the LANDSCAPE button.

2. Always try to fit UNBROKEN tables onto a single page for printing: that is, if your table is too long for 8.5" margins, it will break into two parts, and you should use LANDSCAPE printing instead.

Replication by Region

1. Most of your cross tabs will be for two variables.

2. Use replication by region (variable 200) only for testing replication of your major finding(s).

Web Presentation

1. To put a large table in web format, click the table, copy with control-c, open a graphics editor, PASTE the image of the table into the editor, then SAVE as in *.GIF format, say as IMAGE.GIF (note: file names are sensitibve to upper/low case in html, the HyperText Markup Language). Then open a file in your word processor and type IMG SRC="IMAGE.GIF" INSIDE ANGLE BRACKETS (I cannot show them here since this IS html) and add text. This image will appear in the web page. I can upload your files to the web, or you can put them in your UC on-line disk space.