Modelling and Analytics
our clients paint convincing stories for stakeholders.
Looking at tabulated data, it often looks like everything is related to everything else. Wallis applies various multivariate techniques to help clients better understand important underlying relationships, and paint more convincing stories for their stakeholders. For example, we use linear and logistic regression models to understand which relationships are important, and which are just noise. We use structural equation modelling to test hypotheses of more complex series of relationships.
Weighting and estimation
Some results from survey data are one thing. Confidently extrapolating results to a broader population is another thing entirely. Wallis helps our clients do this with confidence. We use various weighting techniques to adjust for bias in sampling/response, and can use modelling to explore potential effects of non-response bias. Our work in this area has been used widely for reports published by all levels of government, and academic literature.
Mining meaning from text
Wallis is adept at uncovering themes from open-ended text. As well as a human coding team, we make use of advanced statistical techniques to mine the relationships and themes emerging from free text. This can be of great benefit: allowing us to hear from respondents in their own words, unconstrained by the themes and structures of closed surveys.
Designing better measures
Wallis can use statistical techniques such as factor analysis to examine the way existing survey items are performing, and confirming their validity. From here, we can advise on changes to questionnaires in order to better capture the concepts that are required, whilst minimising the burden on respondents.