11. Reporting Results
Reports based on survey data should give the reader enough information to appreciate most of the important elements, as described below.
Reports should clearly describe the scientific methods that were applied and any potential biases associated with these methods, such as:
- A clear definition of the target population and the variables being measured;
- A clear statement of how the sample was selected, how respondents were contacted, and parts of the population that are not included in the sampling frame or the responding units;
- Disposition rates such as contact, cooperation, and response rates for the sample and possibly for important subpopulations;
- Response rates to important questions, such as questions related to income or negative behaviors such as drug use, particularly if missingness rates for the question is higher than unit nonresponse from the survey; and
- Any other substantial problems with the survey.
In terms of numerical estimates, one should report estimates along with standard errors (or margins of errors, confidence intervals, or mean square errors) and sample sizes. If one reports hypothesis tests, then ideally estimates, standard errors, and p-values are also reported. Simply reporting a p-value does not effectively communicate much useful information. Professional statistical standards should be followed when forming tables and figures. And, of course, it is critical to interpret the resulting estimates in the context of the scientific problem being studied.