Coding Open Ended Questions
- classifying answers / converting to numbers
- open ended
o attribute information – range of answers is too large
o attitudinal – response options are unknown / feedback is required.
o general feelings
o reasons for opinions
- other category
- tradeoff : more codes more detail / less codes easier the analysis
- codes
o pre-existing
§ systematic / developed by experts
§ publically available / coding transparent
§ same coding for repeated surveys
§ facilitate comparisons
o developed based on responses
§ read selection of responses
§ summarise responses into themes
§ if required – group themes into broad topics
§ generate frequency distribution for each theme
Thematic Coding
Coding Missing Data
- different from valid code
- reasons
o not required to answer
o not ascertained
o refused to answer
o did not know answer / no opinion
Checking for Coding Error
- sources of error
o data entered in wrong column
o miscoding
§ data collection
§ manual coding
§ data entry
- methods for checking coding errors
o valid range checks
o filter checks
o logical checks
Preparing variables For Analysis
- Changing categories
o initial coding results in more categories than we require
§ recode occupational categories into white / blue collar
o too few subjects in some categories
o collapsing categories can highlight patterns in data (but can also mask a relationship)
o approaches
§ substantive
· combining categories that have something in common
o industry based categories
o amount of training
· divide categories of variables into equal lots [gw ?]
§ distributional
· restricted to ordinal and interval variables
· divide sample into roughly equal sized groups of cases
- rearranging categories
o arrange categories in more logical order
§ more appropriate to focus of analysis
§ tables easier to read
§ changing level of measurement of variable and thus affecting the methods of analysis that can be applied to variable
o example
§ organize industry categories according to level of unionization
- reverse coding
o when constructing scales
o change direct of scale to be consistent
Creating New Variables
- create new variables
o developing scales
o conditional transformations
§ eg, marital history of both husband and wife
o arithmetic transformations
§ age difference between husband and wide
Standardising Variables
- interested in scores relative to other people in sample
- comparable studies where units of measure are not comparable (eg, income) ??
- remove inflation
- interval level à z scores
- ordinal level à percentiles
Dealing with missing data
- checking for missing data bias
o divide sample into 2 groups based on whether particular variable is missing data or not
o cross tab
- methods for dealing with missing data
o deleting either cases or variables
§ list wise deletion
· any case with missing data deleted
· issues
o loss of data / reduction in sample size
§ pair wise deletion
· use only cases with complete data for each calculation
§ deletion of variable
o statistical imputation
§ sample means
· value of mean of that variable
§ group means
· divide sample into groups on background variable
· issue
o exaggerates extent to which people in a group are similar
o inflates correlation between variables
§ random assignment within groups
· divide sample into groups on background variable
· replace missing value with value of same variable of nearest preceeding case
· maintains variability
§ regression analysis
Gw comment
- is a relationship being masked as a result of collapsing same / similar to Simpsons law
No comments:
Post a Comment