Please enable JavaScript.
Coggle requires JavaScript to display documents.
Weighting and preparing data (Computer-based data collection (inconsistent…
Weighting and preparing data
Weighting
statistically disadvantage
cure non representative sample
nonresponse
weighting: better estimate
no effect on estimates if no association between variable and characteristics
improves estimates if nonrespondents are similar to respondents
not recommend to adjust known characteristics
assumes response is the same
Nonresponse
include
ill
refuse
no language
unavailable
exclude
wrong address
dead
moved
survey not delivered
non-random nonresponse
not same probability
related to characteristics
relationship between non response and:
topic
interview mode
respondent/interviewing characteristic
bias and hard to measure
random nonresponse
ineffecient
no relationship > no bias
Data reduction (recoding)
Coding (categorizing)
Data entry
Designing code
Data cleaning (completeness)
Entry errors
coding decision errors, misapplication of rules
transcription error when recoding answer/number
Formatting data file
code data in order of survey
always put entry
serial identifier --> checking file completeness
Constructing codes
ambiguous code = unreliable and uninterpretable
Common principles
consistent assigning of numbers
codes fit real life (years, hrs)
include missing data
inapplicable info
'don't know'
not ascertained info
refused to answer
answers into numbers
Approaches to coding and data entry
data entry
programs
organized
quality control
checks consistency
handles contingency questions appropiately
only legal codes
one-step process: CATI, CAPI, CASI
two-step: 100% verified
interviewer coding
interviewer decisions coding minimal
if they do: verbatim
quality control
train
supervize
notes if not sure
Computer-based data collection
inconsistent data: flagged and alterned
takes previous info into account
downside: time for testing and debugging + no quality control for data entry
follow complex Q patterns
immediate analysis after adding data
Data cleaning
error > consult original source
check again
no illegal codes
when done at data entry: less reliance on cleaning
make complete and in order
Reliability of coding open Q's
quality of coder
training
quality of question
supervision
Getting more participants is not efficient, therefore use weighting
Data preparation
Closed Q's: error less than 1%