Triangulation

The text on this page is taken from an equivalent page of the IEHIAS-project.

Triangulation is the process of making repeated estimates of the same phenomena, using different (and independent) methods. Its aim is not to validate any specific method, but to give confidence in the final results (and obtain an indication of their possible uncertainties) by drawing on different approaches.

To be informative, it is essential that the methods used are truly independent – i.e. are based on unrelated principles, data sets and/or systems of analysis. One well-established example of its use is the production of so-called ensemble forecasts of the weather, made by running different meteorological models and then pooling the results.

Triangulation has special value in relation to screening, for the simple and rapid methods of assessment that have to be employed for this purpose often have limited credibility, and may be subject to significant errors. Examples might therefore include:

• Using a range of different survey methods (e.g. focus groups, questionnaires, delphi studies) to elicit expert opinions on potential health impacts of a policy;
• Using expert elicitation and extrapolation techniques to estimate potential impacts of a climate change scenario;
• Using different proxies (e.g. based on distance from source, source intensity) to estimate exposure distributions across a study population.

The estimates made by these various methods might then be compared, to indicate the likely range of uncertainty inherent in the preferred estimate, or they might be averaged to provide a pooled estimate. In the latter case, it may be appropriate to weight the results, to reflect their relative uncertainties, and it is certainly likely to be helpful to report the range of values obtained (or some measure of their variability).