# Survey data analysis: independent variable ordinal, dependent variable dichotomous, various complications...

#### GuyNorman

##### New Member
Can anyone suggest an appropriate analysis for the following situation? Simplifying somewhat, we interviewed 1000 respondents living in slums in Nairobi. We gave them a list of 17 basic services (healthcare, water supply, etc...) and asked them to rate current service provision level for each of the 17 services, on a 4-point scale (0= non-existent, 1 = poor, 2 = adequate, 3 = excellent). We also asked them to identify their top 5 priority services for improvement. In this analysis, we want to assess whether prioritisation is influenced by (or more strictly, associated with) perceived current service level. So for each respondent we have a 4x2 table, with 4 rows Service Level (ordinal: 0, 1, 2 or 3) and 2 columns Prioritisation (dichotomous: top-5-priority Yes or No); the table has a total of 17 values (of which 5 forced Prioritisation = Yes, 12 forced Prioritisation = No). Service Level can be considered the independent variable, Prioritisation the dependent variable. I think (???) we need some sort of ANOVA analogue, maybe considering the data for all 1000 respondents and all 17 services together, and maybe with Factor “Respondent” nested within Factor “Service Level”? [Or for all 1000 respondents but considering only one of the 17 services at a time.] Friedman Test??? Factorial Logistic Regression??? GLM??? No idea! Advice appreciated! Thank you! [Please assume for the purposes of this initial question that the 1000 respondents are representative of "slumdwellers in Nairobi", though in reality it's not so straightforward.]