Mean comparison: column significativity (AskiaAnalyse)
You should use column significativity (for numeric variables) when you want to compare means between two independent samples. In the table, letters such as “a", "A" or "A+" are displayed to indicate the level of significance. For example:
To define where the letters in the header will be displayed, in the general tab of table definition, click settings... next to tab template, and open the total and caption tab. Select as desired from show a column order letter (for col significativity), show a column order letter in edge total column, show a column order letter in total column, show a row order letter (for row significativity).
Once you have defined where the letters will appear, select the calculation column significativity or row significativity in the general tab of table definition:
Note that the calculation column significativity for numerics is found in the section for numeric questions (above Z-Score (numeric)).
To define the parameters of the test, in the general tab right-click column significativity, select properties..., then click advanced options.... Then, set the options as follows:
Let’s imagine we want to compare the Overall Rating given for a movie by Social Category X Gender:
The available options vary depending on the type of calculation being defined. In our example, the following options are available.
- Test each column against: Select what comparison will be made, as follows:
- All columns: all the columns of questions selected in columns (independently of the question) will be compared with each other. For example, we will compare A/B, ….C/G, …H/K.
- All columns of the question: All columns of the corresponding question will be compared with each other. We will compare only between the Age categories (C/D) and between the CSP categories (E/F).
- The total only: Only the column Total (A) will be compared to the other columns.
- All columns of the question and the corresponding responses edges: all the columns of the question will be compared with each other and with the corresponding edge. For example, we will compare C/D or E/F or C/H.
- Previous column only: all the columns of the same question or different questions will be compared to the previous column.
- All columns of the question and the corresponding questions edges: all the columns of the same question (only) will be compared with the corresponding edge. For example, we will compare C/H or D/I or E/J but we will not compare C/D or E/F.
- All columns of edge responses: Only the columns within the edge response will be compared. For example, between B-C-D-E-F or between G-H-I-J-K.
- All columns of the current question and the first question: All the columns of the first question will be compared to all columns of the current question.
- Specify columns: You can choose the columns to compare (see
test cols option below).
- Show letter in: Select the priority for the display of letters, as follows:
- Both columns: the significance will appear in both columns.
- First column: the significance will appear in the first column.
- Column with highest value: the significance will appear in the column where the column profile is highest.
- Using: Specifies the type of test to be used (for more information on these test types, follow the links below):
- Display minus if under: When this option is selected, a minus sign is shown if the significancy goes down. This can be used in conditional formatting when you test in conjunction with the column before: if there are two letters, the value has significantly gone down, if you have one, it has gone up.
- Use student test when degrees of freedom <: If this option is selected, a student test will be used when the degrees of freedom are less than the amount stated in the adjacent box.
- Use unweighted base: If this option is selected, column significativity will be carried out on weighted mean and unweighted counts.
- Columns are assumed independent: When this option is selected, the individuals belonging to a sub-total will be considered different to those present in the category grouped in the same sub-total.
- Count threshold: The minimum count to be taken into account in a cell.
- Base threshold: The minimum base that must be met before column sig. testing is displayed. By default, the minimum base is 0. This option affects numeric questions.
- High significativity (%): The percentage at which values are to be regarded as highly significant.
- Normal significativity (%): The percentage at which values are to be regarded as of normal significance.
- Low significativity (%): The percentage at which values are to be regarded as of low significance.
- Display "A+": Mark highly significant values with A+.
- Display "A": Mark values of medium significance with A.
- Display "a": Mark values of low significance with a.
- Test cols (A:B,C:D-F): This option appears when you select
specify columns in test each column against. It allows you to list specific columns to be compared against each other.
Enter the letters or numbers of the columns you want to test. Separate each test with a comma, and use a colon to separate the columns to be compared within a test. Examples:
- To test A vs B and D vs E only, enter A:B,D:E or 1:2,4:5
- To test A against B,C,D,E,F: enter A:B-F
- To test A,B,C against A,B,C: enter A-C:A-C or A-C
You can display the list of columns being compared in your table (e.g. the footer), by entering the keyword ??ColSig?? in the appropriate field.
- Sig message (use ??sig??): Defines a message which indicates at which level the columns have been tested during column significativity testing. The token ??sig?? can then be used to display this message. Select a message in the drop-down list, or enter your own. The items you can place in the message are:
- p1 = high %
- p2 = normal %
- p3 = low %
- invp1 = 100 - high %
- invp2 = 100 - normal %
- invp3 = 100 - low %
- p1_1 = high proba ( 0-1)
- p1_2 = normal proba ( 0-1)
- p1_3 = low proba ( 0-1)
- invp1_1 = 1 - high proba ( 0-1)
- invp1_2 = 1- normal proba ( 0-1)
- invp1_3 = 1- low proba ( 0-1)
For example, if the message is "Columns are tested at ??p1??", p1 will be replaced by the high significativity value.
For information on the following column significativity types, follow the links below: