when a respondent is unable to answer a question because it is distribution of data (counts) for the other responses, but it outside the scope of their role or experience. For this reason, we does change the size of the denominator used to calculate the frequently add a “Don’t know or not sure” (DKNS) response to the percentage of responses across the remaining responses. This has list of responses for a question. However, this creates a conun- the effect of proportionally increasing the percentage values of the drum regarding what to do with DKNS responses. remaining responses relative to the number of DKNS responses. The number of valid cases is adjusted accordingly. Where we One approach is to treat it just like any other response. In this have elected to exclude DKNS data, a careful examination of the way, report readers can see the percentage of respondents that footnote for the figure will enable the reader to determine the answered DKNS. The advantage of this approach is that it reports number of DKNS responses based on the difference between the back the exact distribution of the data collected. The challenge sample size (DKNS included) and valid cases (DKNS excluded). with this approach is that it distorts the distribution of valid responses—those responses where respondents could answer the Finally, percentage values in this report may not total exactly 100% question. due to rounding. Some of the analyses in this report excluded the DKNS responses. This can be done because the data missing can be classified as either missing at random or missing completely at random. Excluding DKNS data from a question does not change the THE 2023 STATE OF OPEN STANDARDS EMPIRICAL RESEARCH ON THE TRANSITION TO OPEN STANDARDS 34