Presentation on theme: "Analysis Plan Regional Meeting of UN Washington City Group on Disability Statistics Rio de Janeiro, Brazil September 19-20."— Presentation transcript:
Analysis Plan Regional Meeting of UN Washington City Group on Disability Statistics Rio de Janeiro, Brazil September 19-20
Objective To test the consistency of the Washington Group (WG) census questions in regards to how their interpretation may differ by: Different core domains Countries Subpopulations
Step 1: Construct Different Measures of Disability For ease of analysis we need binary variables To not lose richness of responses, we can used different definitions D1=1 if some difficulty, a lot, or can’t do at all D2=1 if a lot of difficulty or can’t do at all D3=1 if can’t do at all
Step 2: Compute Disability Prevalence Prevalence measured four ways: P1=1 if at least one domain has D1=1 P2=1 if at least one domain has D2=1 P3=1 if at least one domain has D3=1 PM=1 if more than one domain has D1=2 These prevalence rates can be compared with measures from previous country instruments to get a sense of what they are measuring relative to the WG questions
Step 3: Construct Extended Measures Extended measures get at a deeper sense of how people are answering the core questions For example, for vision: Core question: Do you have difficulty seeing, even if wearing glasses? Extended questions: A) Do you have difficulty seeing and recognizing a person across the road? B) Do you have difficulty seeing and recognizing an object at arm’s length?
Question Order In order not to “contaminate” the results of the WG questions, it is important that all six core questions be asked prior to the extended questions
Constructing Extended Measures Within each domain… ED1 = 1 if at least one extended question in that domain has a response of some difficulty, a lot of difficulty, or can’t do at all ED2 = 1 if at least one extended question in that domain has a response of a lot of difficulty or can’t do at all ED3 = 1 if at least one extended question in that domain has a response of can’t do at all
Step 4: Correlation analysis When people are answering the core questions, how does that relate to the underlying concepts in the extended questions? It could be that people respond positively to one set or questions but not the other, so define D0 = 1 if there are no reported difficulties ED0 =1 if there are no reported difficulties in any of the extended questions
How do core and extended questions relate? Using correlation coefficients test if there are statistically significant differences between the D and ED measures D0 vs. ED0, D1 vs. ED1, etc. If they differ, then the measures are not equivalent If they aren’t equivalent, does one set identify more people as disabled?
Step 5: Differences by Subpopulation Older people Gender Rural vs. Urban Ability to examine subpopulations depends on sample size.