Note – different weights calculations crosstabs & custom tables
Hi, I got a question from a customer why the crosstable and custom table showed different result, when they worked with weighted data. Because of the weighting, there seems to be different types of roundings of the frequencies and therefor the calculations differ with one decimal. One decimal could be important in some situations, so let me explain the differences.
This is my example of data file, and you can see the weight variable to the right (w1).
I have compared the 2 different commands: Analyze – Descriptive Statistics – crosstabs and Analyze – Custom Tables – Tables
Result: Here are the 2 tables and I have marked with yellow when it differs:
The reason is how the absolute frequencies (the Counts) are rounded, and you can choose between different types in the Crosstabs-command. This is the default setting (see the “Cell” subcommand within the Crosstabs-comman):
Here you can see the 2 tables Counts that is the reason for different percentages.
As you see the upper crosstabs shows no decimals (it’s rounded to the nearest integer):
If I instead choose “No adjustments” instead in the “Cell”-subcommand in crosstabs, the result will be different.
Result after this setting, and now you can compare to the Custom Tables above and
see that the yellow marked percentages are exactly the same in both tables:
I have been in Contact with the developer’s site where you can wish updates in coming versions of SPSS Statistics, where I wished that you should have the same default settings in both crosstabs and custom tables. But I can also understand that some user rather want to round the Counts to integer (crosstabs default setting), so it’s hard to get all user satisfied.
The important thing right now is that you are aware of the differences.