Chapter 12 Flashcards
name the 5 steps in the data validation process
- Validate
- Edit
- Code
- Capture (MS Excel)
- Import into analysis package
(Qualtrics)
Name the 5 steps of the validation of data guidelines
- Check-backs
- Review questionnaire & interviewing instructions
- Evaluate interviewer reputation
- Field editing
- Central office editing
Explain check backs
- Check-backs:
*Telephone 10 – 20% of respondents named on questionnaire to
check if they were interviewed.
*Randomly selected
*Respondent questioned & answers compared.
*Respondent comments on interviewer manner & behaviour
Explain steps 2-3 from the validation of data guidelines
- Review Questionnaire & Interviewing instructions
*Does the respondent meet the sample requirements?
e.g. Sample requires females to be interviewed, but a questionnaire was
completed by a male - Evaluate Interviewer reputation
*Inspect call sheet; Interview conducted according to correct
sampling procedures?
e.g. Interviewer was supposed to use systematic sampling but used
convenience sampling.
Explain editing
- Editing
*Check correctness & completeness
*Adjust questionnaires where necessary & discard worthless
forms
*Field & central office editing
What does field editing and central office editing concern?
(1)Field editing: -Preliminary editing done by field organiser
-Detects obvious omissions and inaccuracies
-Useful method for resolving any misunderstandings in fieldwork procedures at an early stage.
+Must be done as soon as possible.
(2)Central office editing:-More thorough
-Done by head office
-Preferably done by a single expert in the field of concern
-Researcher decides :(1) What to do with gathered data,(2)handle incomplete questionnaires, (3) Do with questionnaires that show lack of interest from respondent
What are the 8 Editing criteria ?
Editing Criteria
*Cheating by interviewer
*Compliance with sampling requirements
*Relevance of answers (misinterpretation of questions by respondent)
*Completeness
*Comprehensiveness & unambiguity of answers
*Comprehensibility
*Legibility & clarity of handwriting
*Inconsistencies
How to handle unsatisfactory questionnaires
*Review quality of questionnaires & interviewers
*Go back to interviewer or respondent for more satisfactory
responses
*Discard unsatisfactory parts of questionnaire
*Discard unsatisfactory questionnaires
What are the 3 coding steps ?
Step 1: The specification of categories;
Step 2: Allocation of code-numbers to each category;
Step 3: Compilation of a codebook or manual
Explain step 1 of coding
Specification of Categories /
Classes
Requirements for specification of categories/classes:
1. Categories should be a suitable size
2. Categories must be mutually exclusive & incompatible
3. Categories must be exhaustive and comprehensive.
Explain step 2 of coding
Allocation of Code Numbers
1. Pre-coding (BEFORE data gathering)
Predetermined categories, code numbers and field positions
Closed questions (Dichotomous & MCQ)
- Post-coding (AFTER data gathering)
Open-ended questions
Review sample of 20% & specify categories in which responses can be placed
Explain step 3 of coding
: Compilation of Code Book
*General instructions on how each variable/question is coded
*Describes each variable, its code name, where variables are placed in computer
record (column/row) & how these must be read (coding).
*No standard procedures for compiling codebook.
What are the issues to consider with Web-Based Questionnaires & Database Data
*Unique respondents (IP address)
*Internet access limits = bias
*Can be combined with paper-based questionnaires
*Cut-off date
Explain verification and cleaning as well as the coding and capturing errors
Verify & clean data IN statistical package (SPSS) since errors may occur
when entering data into computer.
Coding & capturing errors:
*Values cannot be executed
*Data transposition
*Foreign data
*Same value entered more than once
*Missing/omitted data
*Records not in sequence
Explain checking for errors
Cleaning tasks done by statistical package/computer programme:
*Checking each variable/question
*Inconsistent and contradictory responses
*Extreme answers
*Missing data
Checking for errors
1 Missing values
-Incomplete fields
-Leave response open, re-interview respondent or remove entire questionnaire
2 Recoding
-Can help rectify errors
3 Labelling of data
- Allocate variable & value labels so data is readable & meaningful
