Answers

1. How unusual is a significant index split?

   About 40% of the healthy standardization sample showed one. Significance and rarity are different questions and the two lines sit three times apart.

2. How many low scores does a healthy person have?

   Across ten measures, 31.6% of cognitively normal adults have at least one at or below 1.5 SD. What counts as unusual depends on how many tests you gave.

3. Is subtest profile analysis worth doing?

   Retest stability of a subtest profile is around .26. We built it, then removed it outright rather than caveating it.

4. How do you compare scores from different instruments?

   An index of 87, a scaled score of 6, a z of -1.5 and a T of 68 are largely one statement in four vocabularies. The conversions, and the two places they go wrong.

5. Why does averaging a T score with an index cancel out?

   Because high is good on one and bad on the other. Direction has to be a property of the measure, or a real deficit quietly averages away.

6. Is a composite double-counted with its own subtests?

   Usually, yes — and the weighting is decided by how many parts the composite happens to have. What to fold out, and where this is still incomplete.

7. What can a performance validity measure establish?

   Less than software usually implies. A gate nobody administered reads as not established, never as passed, and nothing is ever labeled invalid.

8. Why compare against your own cases at all?

   No published norm table describes the people who come through your door. A worked comparison out of the demonstration, with the confounds reported rather than corrected away.

9. Which identifiers does Casebook store?

   One direct identifier — a date of birth, if you enter it — and no column for any other. What that means, and why it is PHI kept under an agreement.

Why these exist

Casebook does not score and it does not diagnose. It holds the scores you already have and describes what a pattern is, with the rule, the comparisons, the scores underneath, what it could not determine, and where the threshold was published.

Three of the first four readings built had to be weakened once the literature was read properly, and one was deleted. Those corrections are the part worth having, so they are written down here rather than left in a commit message.