Exploring the Theoretical Basis for Stratifying Scale Results

Some scales in practice (or online) divide numerical scores into levels (e.g., low, medium, high) to facilitate understanding of results. Such classification can lead to inaccurate measurement for subjects at critical values; for example, individuals differing by 1 point between categories may not differ more than individuals within the same category differing by 2 points. Although classification has played or still plays an important role in the history of psychology (and in the dissemination of psychology), this discrepancy cannot be ignored when accurately interpreting psychological data.

Of course, the way data is classified also affects the classification method. In my personal understanding, mapping specific scores onto a population distribution curve and using standard deviation as a measure might be a good approach (i.e., similar to percentile ranks in the population). However, the comparison population needs to be studied, and the difficulty for readers to understand will also increase. If the mean of the measured variable in society increases during a certain period (e.g., an increase in disease prevalence), should the diagnostic threshold be raised accordingly? It seems inappropriate.

According to mainstream understanding, comparing multiple similar measurements might be a good method, at least allowing mutual reference (External Validity). This is also a way to more comprehensively understand the subject. Ultimately, however, how to determine the relationship between scale measurements and unmeasurable definitions (i.e., Validity) remains a question worth studying.

If there is an opportunity, I will consult the literature for further in-depth study.)