In some cases, non-categorical numerical variables in a dataset contain values that are used for indicating non-response or for top/bottom coding.
An example of this would be an AGE variable, whereby values like -9 may be reserved to indicate that the respondent did not want to answer the question, or a value of 99 to capture individuals aged 99 of higher. When that is the case, these values need to be treated differently for certain operations. When for example computing the average age, the -9 value should be excluded as it would bias the results.
To properly handle these for computation and aggregations purposes, RDS supports configuring “reserved values” on a variable (sometimes also referred to as “sentinel” values”).
A reserved value is by default treated as a user missing value, which as the following effects:
it is excluded from valid counts and summary statistics
It is excluded from tabulation using statistical function (e.g. sum, mean, avg)
It remains included in record counts
In the example above administrators may want to treat a non-missing reserved value, such as 99, as a valid value to include in statistical functions. To allow this distinction, reserved values can be flagged to include them in computations. This flag is set to false by default.
Reserved values can be set as variable properties through the API and the RDS Manager application.
Note that when importing metadata from a DDI-Codebook file, if all categories on a numeric variable (var/catgry elements) are flagged as missing, these will be treated as reserved values (instead of a classification). Such variable will be flagged as a measure (can be used with aggregation functions) as well as a dimension (can be used for tabulation) as long as it has the reasonable number of distinct values (configured by the server administrator, the default being 500). Note that this excluded variables holding non-integral values.
Supporting and understanding how to handle reserved value is power features that makes RDS a unique platform for delivering intelligent data services to users.