Numeric metrics: Vol, Mean, Median, Min, Max, SOV

Numeric metrics work with the values of respondents rather than with how many respondents there are: the sum of values (Vol), their average (Mean), the middle value (Median), the extremes (Min, Max), and the share of the total value (Vert SOV, Horiz SOV).

Which values are used

For every respondent in a cell, DataTile takes one value:

• Numeric variable: the respondent's answer, for example age 34 or 3 visits.

• Categorical variable: the digit (midpoint) of the respondent's category. Every category gets a digit when the data is loaded, usually its code or its position in the list; you can change it in the Meta-Editor, see Midpoints: digit values for categorical variables. For an income band "1,000–2,000" a digit of 1,500 gives a meaningful average; the default sequential digits 1, 2, 3 usually do not.

• Variable in the VALUE drop zone: if you put a numeric variable there, its values are used for every cell, whatever the rows and columns are. See Value drop zone in the crosstab area.

By default values come from the row variable. If the rows are categorical and the columns are numeric, values come from the columns. The selector in the VALUE area lets you choose rows or columns explicitly.

A row that is a single category gives every respondent in it the same digit, so its Mean, Min and Max all equal that digit. To average a scale per group, put the scale in the VALUE drop zone and the groups in rows and columns.

Respondents without a value are left out of all numeric metrics.

Vol

Vol (vol), volume, is the sum of the values in the cell, each multiplied by the respondent's weight:

Vol = Σ weight × value

Count is the same sum with 1 in place of the value. Use Vol for totals such as total spend, total number of purchases or total minutes watched.

Mean

Mean (avg) is the weighted average value:

Mean = Vol ÷ weighted number of respondents in the cell with a value

Mean is shown with one decimal by default. It supports significance testing, see Standard T-test in DataTile. Standard deviation is not offered as a separate metric; DataTile calculates it internally for that test.

Median

Median (med) is the value that splits the respondents in the cell into two halves by weight. Values are sorted, their weights accumulated, and the median is the value at which the accumulated weight reaches half of the total weight.

For a categorical variable DataTile treats each category as an interval around its digit and interpolates inside the category where the halfway point falls. Each category's interval ends halfway to the next category's digit. The first category starts at its own digit and the last category ends at its own digit. So the median of a categorical variable can fall between two digits.

Median is off by default and shown with one decimal.

Min and Max

Min (min) and Max (max) are the smallest and largest value in the cell. Weights do not affect them. Both are off by default.

Vert SOV and Horiz SOV

SOV means share of volume (originally "share of voice"). It is the volume counterpart of Vert% and Horiz%:

• Vert SOV (vsv) = Vol of the cell ÷ Vol of the whole column × 100. The column volume is the sum of the row variable's values of everyone in the column base.

• Horiz SOV (hsv) = Vol of the cell ÷ Vol of the whole row × 100. The row volume is the row's Vol across everyone in the base of the column variable.

Example. "How many cars does your household have?" has categories 0, 1, 2 and 3 with matching digits. Among 50 men:

30% of men live in two-car households, and those households own 46% of all the men's cars.

Special multi-response sets

Two MR-set types change what Vert% shows, because a count of respondents means little for them:

• Quantitative MR-set: Vert% shows the share of volume, like Vert SOV. Use Mean for these sets. See Quantitative Multi-response Set.

• Mean MR-set: Vert% shows the mean as a percentage of the declared maximum. See Multi-response Set for Average aggregations.