More obvious, glaring problems with the Meyers-Briggs:
1. Score distributions along each axis are normal, not bimodal.
2. Scores are discretized to one-bit values along each axis.
3. Error bars are big enough that the combination of facts (1) and (2) results in very similar people getting very different scores, and poor test-retest reliability.
Though my personality has changed over the years, it has stayed similar in a lot of ways. The person I was at 5 has a great deal of overlap with the person I am now. And all versions of me who know about probability distributions would be unanimous in their mathematical disapproval of how the raw test scores from the Meyers-Briggs get reported and interpreted.
1. Score distributions along each axis are normal, not bimodal.
2. Scores are discretized to one-bit values along each axis.
3. Error bars are big enough that the combination of facts (1) and (2) results in very similar people getting very different scores, and poor test-retest reliability.
Though my personality has changed over the years, it has stayed similar in a lot of ways. The person I was at 5 has a great deal of overlap with the person I am now. And all versions of me who know about probability distributions would be unanimous in their mathematical disapproval of how the raw test scores from the Meyers-Briggs get reported and interpreted.