This test was administered at the end of the survey for 600 of the responders and at the beginning for 124 responders to test for survey fatigue.
As a solution, we present a test for differences in distribution based on an existing test for survey data based upon randomizing the data at the level of individuals rather than observations.
Rather than test for survey understanding, these tests are included to investigate whether responses are 'rational' according to the axioms of preference-based consumer theory [ 24].
Safety issues in surveys consider consequences of collection of biological specimens and clinical testing for survey participants and for data collectors.
Two-sample t-tests for survey data and the Rao-Scott chi-square test were used for continuous and categorical data, respectively.
Proportions were compared using Pearson chi-squared tests corrected for survey design and the continuous measures (age and household size) with Wilcoxon rank-sum tests.
In univariate analyses for categorical variables, we conducted Pearson chi-squared tests corrected for survey design with the second-order of Rao and Scott [ 30], and converted the test into a design-based F statistic.
Chi-square tests, adjusted for survey design (using Complex Samples module in SPSS 20, with strata = country and cluster = facility) were used to test significance; p-values of less than 0.05 were considered significant.
To test for differences in proportions in net use or infection prevalence a Pearson χ2 test accounting for survey design (clustering and stratification) was used and the test statistic converted to an F-statistic using the second-order Rao and Scott (1984) correction yielding wider confidence intervals and conservative P-values compared to the uncorrected χ2 test.
The cognitive interviewer who conducted interviews in the Urdu language completed two surveys, one in Urdu, and one in English for all rounds of testing for 18 surveys.
