Analysis of complex data requires advanced survey sampling knowledge and access to special software packages developed for weighted data. Because weighting increases variance of survey estimates, using standard variance calculation formulas with weighted data can result in a number of misleading statistical inferences, such as:
- detecting significant differences when observed differences are actually statistically indistinguishable
- erroneous rejection of null hypotheses
- creation of unreliable models for prediction and forecasting
Our seasoned survey research statisticians have decades of experience working with weighted data from complex surveys using special procedures from virtually all available software packages, including SAS, SPSS, SUDAAN, Stata, and R. We provide you with robust analytical support that include:
- weighted descriptive and bivariate statistics
- advanced multivariate procedures such as regression modeling, contingency table analyses, and various segmentation procedures
Moreover, we offer a range of post-field data enhancement services:
- imputation of missing data
- nonresponse bias analyses and associated weighting adjustments
- creation of final analysis weights
For imputation we use the latest techniques, including different hotdeck and multiple imputation procedures. For weighting we specialize in iterative proportional fitting (raking) and propensity modeling.
Our services also include customized workshops in sampling, weighting, as well as analysis and reporting of survey data.