Confidence intervals in psychological tests

[DrDu]

Recently I saw the results of an IQ test given with confidence intervals. While I know very well what a confidence interval is, I am not sure how you can give one for a psychological test. I think to determine a CI you either have to know the "true IQ" of an ensemble of persons and see how the results of the test scatter. Or you must be able to repeat the test with the same person and see how the results scatter on repetition. However, you neither know the true IQ nor can you repeat the same test with the same person as it will remember previous results.

 

[Monstrous Math]

IQ is an ambiguous term, so it seems doubtful that there is a "true" IQ that is always the same for that person. If the person is very very tired, for example, it can skew the results. Or if the person is on a drug. Or in love. Or whatever. In any case, even if you did know the true IQ, you wouldn't need a confidence interval. Confidence intervals are used because one is uncertain. As for what specific data they used for their confidence interval, you'll have to get that from the researcher. It's hard to discuss data we're in the dark about.

 

[DrDu]

I know that the IQ can vary. My question is not specific to the IQ but how you define confidence intervals in general for psychologic quantifiers where you have problems with repeatability for one subject.

 

[256bits]

The confidence interval would refer to the sample population and NOT the individual. The test score is what the individual received with a margin of error

It describe the score one receives on his/her IQ test of being part a sample and how reflective it is to the whole population.

In other words it is comparing a sample population, of which the testee is part of, to the whole population, and then asking if the testee then does belong to a sample, of all the samples that can be taken of the population, reflecting the parameters of the population, or not.