More than once, I noticed that researchers sum ordinal scores of e.g. items of a psychometric test to gain a total score. This score is e.g. calculated for multiple study participants and metrics like mean or standard deviation are deduced. Isn't this a mistake or am I missing something?