MICS DATA PROCESSING Secondary Editing
Mar 27, 2015
MICS DATA PROCESSING
Secondary Editing
REMEMBER AND REMIND YOUR FIELD STAFF:
The best place to correct data is in the field where the respondent is available to resolve inconsistencies. Once the questionnaires reach the office, the best you can do is to apply consistently fully and carefully specified editing guidelines.
Secondary Editing Flow Chart
Backup Raw Data File
Secondary Editing
Backup Final Data File
Resolve Inconsistencies
Correct Raw Data File
DP Supervisor
DP Supervisor
Secondary Editor
DP Supervisor
DP Supervisor
Inconsistencies?
No
Yes
General Rules for Resolving Inconsistencies
Review all pertinent responses in the questionnaire(s).– For skips check responses preceding and following.
Refer to the editing guidelines Do not make up an answer - if necessary, use
codes for inconsistent or missing Change the fewest pieces of information Leave the inconsistency without correction and
document the inconsistency for users
Data Editing Philosophy
Field Editing– Interviewer or field editor
Using field editing manual can be fully corrected
Office Editing - Use editing guidelines– Office editor
ID and structure errors only
– DE personnel Check for data entry errors; resolve only structural inconsistencies
– Secondary editor Investigate and resolve (sometimes by taking no action) all
inconsistencies
Four Examples
1. Woman’s age and date of birth inconsistent
2. Dates of DPT1 and Polio1 vaccinations different
3. Level of education is inconsistent
4. Date of Polio 3 vaccination before date of polio 1 vaccination
Example 1: Basic Information
The Data– WM6 = 04/2005 = 1264– WM8 = 09/1962 = 753– WM9 = 41
The Error Message U 1003 E Age of woman (WM9=41) and her date of birth
(DOB=09/1962) inconsistent [DOI=04/2005]
Example 1: The Inconsistency
The Inconsistency– Age
calculated age (calcage) = 42 reported age (WM9) = 41
– Date of birth calculated LDOB: 1264 - (12*41) - 11 = 761 calculated UDOB: 1264 - (12*41) = 772 reported DOB: 753
Example 1: Resolving the Inconsistency
Variables to Check– WM6, WM8, WM9, HL5(LN), CM2, MA6
Steps1. Check for data entry errors2. If WM6M = WM8M, and WM9 = calcage - 1,
leave unchanged3. If WM8M and WM8Y valid, set WM9 = calcage4. If WM8M invalid, set WM8Y = 9997 (inconsistent).
Example 2: The Problem
The Data– Polio 2: IM3C = 08/08/2003– DPT2: IM4B = 08/08/2004
The Error MessageU 2705 M Date of Polio 2 vaccination (08/08/2003) and
date of DPT2 vaccination (08/08/2004) different
The Inconsistency– polio and DPT shots are often given on the
same date
Example 2: Other Information
Vaccination dates– Polio 1: IM3B = 16/06/2003– Polio 2: IM3C = 08/08/2003– Polio 3: IM3D = 13/09/2003– DPT1: IM4A = 16/06/2003– DPT2: IM4B = 08/08/2004– DPT3: IM4C = 13/09/2003
Example 2: Resolving the Inconsistency
Steps1. Check for data entry errors2. See if recording mistake was made on
questionnaire3. If no obvious recording mistake, leave data
unchanged.
Example 3: Basic Information
The Data– ED3A = 2 { secondary }– ED3B = 11
The Error Message – U 0090 E ED1=02: Level (ED3A=2) and grade (ED3B=11) of
education inconsistent
The Inconsistency– ED3B records grade at the current level, and for
this country (UK), the highest secondary grade is 7.
Example 3: Other Information
Other Variables– Current schooling: ED6 = notappl– Schooling last year: ED8 = notappl– Highest level (woman’s questionnaire):
WM11 = 2WM12 = 11
Example 3: Resolving the Inconsistency
Steps1. Check for data entry errors2. Check for interviewer errors
a. Does ED3B include grades passed at lower levels?
3. If available, check values of WM11 and WM124. If you can’t resolve inconsistency, set ED3B = 97 (inconsistent).
Example 4: The Problem
The Data– IM3B = 25/11/2003– IM3D = 08/01/2003
The Error MessageU 2704 E Date of Polio 1 vaccination (25/11/2003)
after date of Polio 3 vaccination (08/01/2003)
The Inconsistency– polio 3 vaccination given after polio 1
vaccination
Example 4: Other Information
Vaccination dates– Polio 1: IM3B = 25/11/2003– Polio 2: IM3C = 03/03/2004– Polio 3: IM3D = 05/01/2003– DPT1: IM4A = 25/11/2003– DPT2: IM4B = 05/02/2004– DPT3: IM4C = notappl/notappl/notappl
Example 4: Resolving the Inconsistency
Steps1. Check for data entry errors2. See if recording mistake was made on
questionnaire3. If no obvious recording mistake, set day,
month and year of most inconsistent date to 97, 97 and 9997 respectively
Adding an Edit
Add logic to the data entry application Add message text to the message file Add message to the editing guidelines
Defining the Editing Specifications
Carefully review the questionnaire Define the edits
– What is the possible inconsistency?– How should the inconsistency be handled
during data entry?– How should the inconsistency be handled
during secondary editing?
Editing Guidelines
For each inconsistency:– explain its nature if error message doesn’t
make it clear– explain how to handle the inconsistency
during data entry (if applicable)– explain how to handle the inconsistency
during secondary editing (if applicable)– in resolution explanations, list all related
variables that should be examined
Modifying the Editing Guidelines
Add editing guidelines for your country specific questions added to the MICS questionnaire
Modify the standard guidelines only after careful consideration by subject specialists
Document any changes to the standard guidelines
Ensure that all processing staff use the manual and apply it consistently
REMEMBER AND REMIND YOUR FIELD STAFF:
The best place to correct data is in the field where the respondent is available to resolve inconsistencies. Once the questionnaires reach the office, the best you can do is to consistently apply fully and carefully specified editing guidelines.