Smart Tables are used to conduct tests of statistical significance for the dependent question compared with each independent question. The results of these tests are added to the Report tree, and ranked according to statistical significance.
To use Smart Tables, select Create > Tables > Smart Tables.
How Smart Tables work
- A significance test is conducted on each table formed by crosstabbing the Dependent Question with each Independent Question (see Planned Tests Of Statistical Significance). The function does a test on all the table cells rather than look at or compare individual cells on the table. It's somewhat like asking, "Is there a significant relationship between these questions?" rather than asking if there's a significant relationship between any of the table's cells. A p-Value is calculated for each of the question combinations/tables such that each table gets a single p-Value.
- This p-value collection then has Multiple Comparisons (Post Hoc Testing) applied. The correction depends on the total number of tests conducted (i.e., the number of tables) and on the distribution of the p-values of those tables.
- These significance test results are shown in the table name (see the example below).
- The tables are ordered according to their p-Value, with the most significant results shown at the top.
- The tables are grouped into the following categories (where applicable):
- Significant. These tables have a p-Value that is less than or equal to the Overall significance level (considering any Multiple Comparison Corrections that have been specified for the table's cells). The p-value that equates to a corrected p-value at the specified Overall significance level is shown in brackets in the name of the first Folder (e.g., Significant p <= 0.071). For more information about corrected p-value, see Multiple Comparisons (Post Hoc Testing).
- p-Value could not be computed. For example, tables where there is no variation in one of the questions.
Some general comments about Smart Tables
- Smart Tables orders results according to p-Values. While p-Values are a useful measure of the strength of association between different questions, there are sometimes better measures. Also, sometimes there will be ties, so the actual order in which things are shown should not generally be treated as a result.
- It is possible for there to be a significant relationship between two questions without any specific cell being significant when compared to other cells on the table. For example, comparing age to preferred brands, you can say that the younger someone is the more they tend to like a certain brand. But, at the same time this table may not show that any one specific age group likes a brand so much that they get picked out as being significantly higher than the rest of the sample.
- Statistical significance is not the same as causation.
- It ignores the likelihood that many of the questions are correlated. An alternative approach that resolves this problem but introduces others is using Trees.
Buttons, options, and fields
Dependent question The question that will be used in comparison against each independent question. This comes from the question currently in the blue drop-down.
Available questions The list of all questions available to use as independent questions.
Independent questions The list of questions that will be tested for statistical significance against the dependent question.
Move the selected questions in the Available questions list to the Independent questions list.
Move the selected questions in the Independent questions list to the Available questions list.
Filter drop-downThe filter variable to apply during the significance testing.
Weight drop-down The weight variable to apply during the significance testing.
The Smart Tables output is a set of tables, appended to the Report. The tables form a part of a folder named after the dependent question (with Smart Tables affixed at the end, e.g., Q7. Company currently with: Smart Tables. Within this folder, there are two sub-folders – one called Significant (p <= 0.05), which contains questions that were significantly related to the dependent question, and another showing the insignificant questions. The cut-off p-value is determined by the settings of the Overall significance level and Multiple comparisons method. Within each folder, the tables are ordered according to their p-values, with the most significant results shown at the top.