Reliability refers to the consistency of measurement. High reliability means that the measured results are highly reliable and will not fluctuate greatly because of the changes of time, situation, personnel and test questions. Including evaluator consistency, repeated measurement consistency and topic internal consistency.

Evaluator reliability is mainly used to examine whether different evaluators have the same score on the same evaluation object; Repeated measurement consistency is mainly used to examine whether the results of a test are consistent at different times and in different environments; Internal consistency is mainly used to test whether the structure of different questions in a test is the same (this concept has been greatly challenged in recent years, and item analysis theory holds that each question measures an independent structure, but it is still very important in classical psychometric theory).

Validity refers to whether the measurement results really reflect what you want to measure. Different from reliability, it emphasizes the practical application value of measurement, but it is based on high reliability, otherwise the measurement results you get are worthless, let alone applied. But if the reliability is high, the validity may not be high. For example, if you take a well-designed eating disorder test to measure depression, even if the reliability is high, its effectiveness is low. It can be seen that validity is often related to how you apply the test results, and also to your interpretation and analysis of the results. If you design a good tool and calculate the score and reliability after measurement, but lock the result in your locker without any form of explanation and analysis, then it is actually invalid.

At present, there are three kinds of validity: content-related validity, standard-related validity and structure-related validity. Two popular validity indicators are predictive validity and structural validity (also called structural validity).

Predictive validity is the degree to which the results obtained by one measurement method can predict the results obtained by another measurement method (standard). It is usually expressed by correlation coefficient. This standard can be a person's performance or a person's performance. Anyway, it is any standard scale. Conceptual validity, which seems to be called structural validity, refers to whether the structure of a measurement method can reflect a certain theoretical structure, such as whether an intelligence test can reflect the four main modules of PASS theory, namely, planning, attention, simultaneous processing and immediate processing. This validity requires factor analysis or other more complicated analysis methods. Predictive validity belongs to the second kind of validity, and conceptual validity belongs to the third kind of validity.