These formats are used to control the significant difference markings on the table.
See also Statistics formats for details of table based statistical calculations available.
See also Significance testing for some general information about significance testing.
See also Column identifiers for details on how to identify columns for comparison.
See also Variable responses for how to add headers and over-headers to group columns for testing.
See also Drill down Distinction mode.
The default levels checked are 95% and 99%.
All tests are two tailed.
All significant testing is comparing figures in a column with figures in another column or the total column.
Mean scores (averages) are always tested using t tests unless format TTV0 is used.
Rows of tables are only tested using Z tests if format SIG is used.
FSL (Footnote for Significance Level)
This logical format causes a footnote to be produced on the table that shows the significance levels tested.
The default value is FSL (true). Use NFSL to turn off.
MCM (Minimum Column for Means)
This format sets the minimum unweighted/ESS number of records in a column for mean score/average comparisons, see format TTV.
The default value is MCM2.
The value must be in the range MCM1 to MCM100.
MCT (Minimum Column Test)
This format sets the minimum number of records in a column for proportion/percentage comparisons, see format SIG.
The default value is MCT30.
The value must be in the range MCT5 to MCT100.
The base value tested will be the base figure for unweighted data. With weighted tables:
With format ESS (default) the effective base will be used.
With format NESS the unweighted base is used.
To see the ESS figures set format ESR.
SHG (Significance Header Group)
This format controls which columns are tested against each other.
The default value is SHG11 (against total and against other columns).
The value must be SHG0, SHG1, SHG2, SHG11, or SH12.
See also format SIG and column identifiers.
SHG0
All columns are tested against the total column, SMA and SMB are used to show significant differences.
NOTE: the actual column tested against will be the total column minus the column in question. This assumes that the records in the column being tested are also in the total column and avoids overlapping samples.
If the default SMA and SMB are used, then cells in the spreadsheet output will be coloured green for significantly higher and red for significantly lower than the rest.
SHG1
All columns under each header are tested against all the other columns under the same header.
IMPORTANT: The columns will only be tested if column identifiers (letters in parentheses) are present at the ends of the columns labels.
Each column is marked with the identifier of all the relevant columns that have a significantly lower mean or percentage than this column.
NOTE: if only lower case letters are used as identifiers then SLA uses lower case and SLB uses upper case.
SHG2
All columns under each over-header are tested against all the other columns under the same over-header.
SHG3
All columns under each super-header are tested against all the other columns under the same super-header (usually all the columns).
IMPORTANT: The columns will only be tested if column identifiers (letters in parentheses) are present at the ends of the columns labels.
Each column is marked with the identifier of all the relevant columns that have a significantly lower mean or percentage than this column.
NOTE: if only lower case letters are used as identifiers then SLA uses lower case and SLB uses upper case.
SHG11 (default)
This does the testing for both SHG0 and SHG1.
Marking will be SMA and SMB as SHG0, followed by identifiers as SHG1.
SHG12
This does the testing for both SHG0 and SHG2.
Marking will be SMA and SMB as SHG0, followed by identifiers as SHG2.
SHG13
This does the testing for both SHG0 and SHG3.
Marking will be SMA and SMB as SHG0, followed by identifiers as SHG3.
SIG (Significance tests)
This format requests Z-tests or t tests that compare percentages in a row with percentages in other cells in the same row.
The default value is SIG0 (not used).
The value must be set in the range SIG0 to SIG6.
Values:
•SIG0 means no testing on individual rows
•SIG1 means Z test with pooled variance
•SIG2 means Z test with separate (unpooled) variances
•SIG3 means t test (recommended)
•SIG4 means t test with continuity correction
See note below about overlapping testing in tables definition:
•SIG5 means overlapping t test
•SIG6 means overlapping t test with continuity correction
Only available in Drill Down:
•SIG7 means discrete t test
•SIG8 means discrete t test with continuity correction
These last 4 tests (SIG5-SIG8) are for breakdowns where a set of columns being compared (under the same header) are multi-coded.
The overlapping test uses a revised formula to take account of the overlaps.
The discrete tests remove the overlapping records from the tests so that the comparison is only between those who are in one, but not the other.
There is a large overhead for these last 4 tests, so they should only be used when it is known that there will be overlapping columns tested.
IMPORTANT: because the tests against the total column are against "the rest" there is no overlap, so:
When comparing with the total column SIG5 will use the formula for SIG3
When comparing with the total column SIG6 will use the formula for SIG4
See also format SHG for the comparisons used.
SLA (Significance Level A)
This is the lowest percentage significance level that will be used to mark cells.
The default value is SLA95.
The value must be set in the range SLA50 to SLA101.
There are two special values: SLA100 means 99.9%, and SLA101 means not used.
See format SMA for the marking used.
SLB (Significance Level B)
This is the next percentage significance level that will be used to mark cells.
The default value is SLB99.
The value must be set in the range SLB50 to SLB101.
There are two special values: SLB100 means 99.9%, and SLB101 means not used.
See format SMB for the marking used.
SLC (Significance Level C)
This is the next percentage significance level that will be used to mark cells.
The default value is SLC101 (not used).
The value must be set in the range SLC50 to SLC101.
There are two special values: SLC100 means 99.9%, and SLC101 means not used.
NOTE: If SLC level is used then an extra + (or - when testing against the total) will be added after the relevant marking used for SLB.
SLD (Significance Level D)
(not recommended, leave as default)
This is the next percentage significance level that will be used to mark cells.
The default value is SLD101 (not used).
The value must be set in the range SLD50 to SLD101.
There are two special values: SLD100 means 99.9%, and SLD101 means not used.
NOTE: If SLD level is used then and extra ++ (or -- when testing against the total) will be added after the relevant marking used for SLB.
SMA (Significance Marker A)
(default is recommended as this will colour cells in Excel)
This character format is the character used to mark cells that are different to the total column at the SLA level.
The default value is SMA'+' (use plus or minus signs to mean significantly higher or lower respectively).
If any character other than + is used then the same character will be shown for significantly higher and lower.
NOTE: Under format MSE the markers refer to this column only (mean differs from zero), not a comparison with the total.
SMB (Significance Marker B)
(default is recommended as this will colour cells in Excel)
This character format is the character used to mark cells that are different to the total column at the SLB level.
The default value is SMB'+' (use plus or minus signs to mean significantly higher or lower respectively).
If any character other than + is used then the same character will be shown for significantly higher and lower.
NOTE: if SMA and SMB are the same character then this will be shown twice for SLB level significance.
NOTE: Under format MSE the markers refer to this column only (mean differs from zero), not a comparison with the total.
TTV (T Test Variance)
This controls the way t tests are calculated on mean scores/averages.
The default value is TTV1.
The value must be set in the range TTV0 to TTV5.
•TTV0 stops testing of means
•TTV1 uses Student's t test (default)
•TTV2 uses separate variances (not recommended)
•TTV5 uses Welch's t test which does not assume equal variances
See note below about overlapping testing in tables definition:
•TTV3 uses an adjusted Student's t test for overlapping samples
Only available in Drill Down:
•TTV4 means discrete t test
These last 2 tests are for breakdowns where a set of columns being compared (under the same header) are multi-coded.
The overlapping test uses a revised formula to take account of the overlaps.
The discrete tests remove the overlapping records from the tests so that the comparison is only between those who are in one, but not the other.
There is a large overhead for these last 2 tests, so they should only be used when it is known that there will be overlapping columns tested.
IMPORTANT: because the tests against the total column are against "the rest" there is no overlap, so:
When comparing with the total column TTV3 will use the formula for TTV1
Overlapping formulae (SIG5, SIG6 and TTV3) use in tables definition
Note that you are only allowed up to 90 columns in a table using these formulae.
Using more than 30 columns will slow down execution considerably. If your breakdown has a lot of responses, then you may want to consider doing the parts that are multi-coded under the same header as separate tables, and only using these formulae on those tables.
Note that the overlapping formulae will not be used if the table is built a column at a time using overlays (grid table).