This underlying issue is often clearer within more complex IDA measurement scenarios. As shown in Table 1, an example from our own work involves three studies that each used slightly different ways of measuring the frequency of alcohol consumption: Study 1 assessed a six-month time frame and responses were open-ended whereas Studies 2 and 3 assessed a 12-month time frame with binned, ordinal response options. Studies 2 and 3, however, each used a different set of frequency bins for the responses. Clearly we cannot simply pool the responses from these three studies given these measurement differences. We can, however, harmonize these items by transforming the original items to have logically equivalent response scales. In this case, some of the response options for assessing the frequency of alcohol use are the same in Studies 2 and 3 and we can collapse other response options to create comparable frequency intervals across studies, thereby obtaining a common set of frequency intervals across the response options for these two studies. For Study 1, we can convert the responses to annualized estimates by multiplying the
