Health Sciences 2801A/B Lecture Notes - Lecture 4: Confidence Interval, Internal Consistency
Document Summary
Internal consistency: concerned with unsystematic error in scores obtained using a multiple indicator instrument. Unsystematic error is a type of measurement error that increases or decreases individual scores by an unpredicted amount. In theory, the indicators that composes an instrument are part of a large, hypothetical collection of indicators that serve to represent a construct. The hypotheti(cid:272)al (cid:272)olle(cid:272)tio(cid:374) of i(cid:374)di(cid:272)ators is (cid:272)alled the (cid:862)ite(cid:373) u(cid:374)i(cid:448)erse(cid:863) I. c. is conceived to the an estimate of the correlation between the existing instrument and the average of all other possible instruments, composed of the same number of indicators, drawn from the same item universe. I(cid:374) a hypotheti(cid:272)al (cid:449)orld (cid:449)ithout (cid:373)easure(cid:373)e(cid:374)t error, a perso(cid:374)"s o(cid:271)ser(cid:448)ed s(cid:272)ore would = his/her true score. In reality, the true score is different than the observed score. The more indicators you have, the more opportunities for this unsystematic error to cancel itself out: a lower internal consistency is predicted when the number of indicators is decreased.