The (!) symbol next to a variable description in a report indicates a low sample size, while (^) indicates an extremely low sample size.
When we model our survey datasets, there are some variables that do not meet the survey’s minimum threshold for sample size. Since it is important to have a certain confidence level in order to use a variable, we drop some of these variables from our final data products. However, we recognize that a dropped variable might leave a gap in your analysis. Our current methodology allows us to include low-sample variables where appropriate for a more granular view of respondents’ attitudes and behaviours. This allows us to preserve the full scale of behaviours so that the distribution can paint a better picture. For instance, “I tend to spend less money on grocery products I buy online than in-store” includes the complete 5-point Likert scale from strongly agree to strongly disagree. If the frequency for the complete scale meets the minimum threshold, we will denote any low sample point in the scale with one of the two symbols.
We would not recommend profiling variables denoted with either of the above low sample flags, but viewing them as part of a range can help interpret the behaviour directionally. These low sample flags also allow us to include variables that might be niche or region-specific, which can help to capture trending themes or market-specific preferences.