Job Settings|Statistics

In Setup|Job Settings, you can specify options for statistics on your tables using the Statistics tab.

Median and quartile options

Sort order - Choose Ascending or Descending as the sort order for Median and Quartile sorting. The default is Ascending.
Type - Choose Interpolated or Medoid as the type of Median and Quartile. The default is Interpolated.

An example will help illustrate the difference between an Interpolated and Medoid median/quartile: Given the data (1,2,3,4), the Interpolated median would be computed to be 2.5, which is the middle value. The Medoid for this same data would be 2, since the Medoid is required to be a value which is present in the data. By definition, the Medoid is selected from the position just before the Interpolated median. Note that if the data is sorted in descending order, the Medoid for the data (4,3,2,1) will be 3. In other words, the same data can have different medoids depending on the sort order. This fact leads some statisticians to prefer the Interpolated median over the Medoid.

Scaling - Allows you to select the number of places to move the decimal point when values are too large to fit in a cell.

Decimal places - Defines the decimal precision for all statistics.

Central tendency - Applies to the Mean, Median and Mode statistics. Select 0, 1, 2 or 3 from the dropdown. The default for Central tendency is 2 decimal places.
Variability - Applies to Standard deviation and Standard error. Select 0, 1, 2 or 3 from the dropdown. The default for Variability is 2 decimal places.

Display options

Show dollar sign ($) - Displays dollar signs for statistics that represent dollars.
Show comma (,) - Displays commas in statistics.

Statistical testing

Means - Displays Means statistical testing if the Means test is also selected on the banner.
Percents - Displays Percents statistical testing if the Percents test is also selected on the banner.

Statistical testing must be selected for a banner AND selected for a table.

Number of lines to use for significance indicators - Specifies the number of lines to display significance indicators for a cell (the intersection of a row and a banner column).
Number of significance indicators per line - Specifies the maximum number of significance indicators to display per line.

Note: Each comparison group (Setup|Banners|Edit Banner|Comparison Groups) can be represented by a significance indicator of up to 3 characters. Each indicator is considered an item for purposes of assigning the Number of significance indicators per line. If the number of significance indicators exceeds the maximum specified and multiple lines of display are not being specified in Number of lines to use for significance indicators, an asterisk will display indicating there are more significance indicators to display. Increase the Number of significance indicators per line or specify multiple lines of display.

Delimiter to use between significance indicators - Specifies the character to use for separating significance indicators within a comparison group.

Note: Using the "/" (slash) character as the Delimiter to use between significance indicators is not recommended. WinCross separates your Comparison groups using this character. It may be difficult to identify your Comparison Groups in the footnotes section of your report with this character also being used as the delimiter.

Significance indicator order

Alphabetically within each comparison group - Significance indicators will be alphabetized within each comparison group. This option is the WinCross default.
Alphabetically across all groups - Significance indicators will be alphabetized across all comparison groups.
Order by significance level - Significance indicators will be listed in order of significance from high to low (for example, A,B,D,f,g,i).

Treatment of statistics

Adjust standard error when weighting - Calculates the standard error in a way that reduces the underestimation of the standard error (and the subsequent overestimation of significant differences) that occurs when treating weights as replications of respondents. This option is considered more appropriate for market research, especially since the market research norm is to use non-integer weights. Adjust standard error when weighting is the default for how statistics are treated when weighting.

Use general formula for standard error (Treat weights as replicates) - Calculates the standard error as simply the standard deviation divided by the square root of the sum of the weights.

Note: When Use general formula for standard error (Treat weights as replicates) is selected and weighting has been applied to your banner, the Means test of Independent (based on test for equal variances) and the Percents test of Independent (using unpooled proportions) are not valid tests for this option.