Validation
Survey Solutions CAPI for surveys/censuses
Nadi, Fiji
Survey Solutions: Validation
Sergiy Radyakin
[email protected]
Development Data Group (DECDG),
The World Bank
March 27-31, 2017
1 / 13
Validation
Data validation
By nature, socio-economic data is commonly bound by some limits,
and answers to questions may be related to each other.
Categorical selections limit values to only the valid ones.
Still one may need to see how different variables agree with each
other.
2 / 13
Validation
Data validation
Validation rules are attached to variables.
Validation consists of up to 10 rules, each consisting of one logical
expression and one error message.
Logical expression must be written in C# language.
Expression may refer to the current variable, or also utilize other
variables in the questionnaire.
Error message is shown when the condition evaluates to false.
3 / 13
Validation
Logical operators
The following logical operators can be used in validation conditions:
Logical operator Description Type of questions
> Greater than Numeric and Categorical one answer
< Less than Numeric and Categorical one answer
== Equal to Text, Numeric and Categorical one answer
!= Not equal to Text, Numeric and Categorical one answer
>= Greater than or equal to Numeric and Categorical one answer
<= Less than or equal to Numeric and Categorical one answer
&& And
|| Or
! Not
Use parentheses as necessary in complex conditions;
To refer to the answer of a question, specify the variable name
corresponding to that question, like so:
age
To compare strings, enclose string constants in quotes, like so:
city=="New York"
4 / 13
Validation
Examples
Verify the numeric variable age is non-negative:
age>=0
Verify the numeric variable age is under 120:
age<=120)
Verify the numeric variable age is between 0 and 120:
(age>=0) && (age<=120)
Verify that the expenditures is no more than income (both numeric
variables):
expendit<=income
Verify that numeric variable age is above 22 if education category
(variable educat) is 4 or 5:
!((educat==4 || educat==5)&&(age<=22))
Validation expressions may be large and complicated. Use parentheses
when in doubt or check the order of evaluation.
5 / 13
Validation
Keyword self
Use the special variable self to refer to the value of the question
itself in validation conditions
(for example for percentages):
(self>=0) && (self<=100)
this makes cloning the questions easier.
6 / 13
Validation
Soft vs Hard validation
Validation is “soft”;
“Hard” validation only for the text questions with a pattern, numeric
questions triggering rosters, numeric integer questions.
7 / 13
Validation
Data validation
C# language provides powerful tools for writing complex validations.
Survey Solutions’ built-in functions greatly simplify some conditions,
and will be covered later on in this course.
8 / 13
Validation
Data validation
The error message must be clear to the interviewer. Make sure all
error messages programmed in the questionnaire are explained to the
interviewers during their training.
When checking coherence of several variables (for example age and
year of birth), one may not be sure which one is wrong. The message
should allow for the possibility that the error is in the other variable.
Make sure validation message is consistent with what is being
checked. If you revise a validation condition, make sure to check if
the corresponding message must also be revised.
9 / 13
Validation
Mandatory questions
Validation validates answers;
No answer, no validation is run;
One cannot make a question erroneous if it is not answered;
Hence no validation condition will make its question mandatory!
10 / 13
Validation
Static text
Static text is an exception;
Always validated;
Can show an error, even though doesn’t carry an answer.
Useful for end-of-section and end-of-questionnaire validations.
11 / 13
Validation
Single-answer enforcement
Instructions like “If child is under 5, enter code 99”;
If there is only one valid answer, often it is best to skip the question;
If necessary, replace the missing value later with a desired constant.
12 / 13
Validation
Final
prevent data entry errors whenever possible, rather than using
validation: use categorical, date, formatted text, etc questions.;
utilize validation for other checks;
test validation: look for false positives, false negatives;
prioritize error messages (do you always want to show all 10)?
13 / 13
