reliability is estimated by the Pearson product-moment correlationcoefficient of two different forms of a measure, usually administeredtogether. Single-administration methods include
split-half
and
. The split-half method treats the two halves of a measure asalternate forms. This "halves reliability" estimate is then stepped up to thefull test length using theSpearman-Brown prediction formula. The mostcommon internal consistency measure isCronbach's alpha, which isusually interpreted as the mean of all possible split-half coefficients.
Each of these estimation methods isn't sensitive to different sources of error and so might not be expected to be equal. Also, reliability is aproperty of the
scores of a measure
rather than the measure itself and arethus said to be
sample dependent
. Reliability estimates from one samplemight differ from those of a second sample (beyond what might beexpected due to sampling variations) if the second sample is drawn from adifferent population because the true reliability is different in this secondpopulation. (This is true of measures of all types--yardsticks mightmeasure houses well yet have poor reliability when used to measure thelengths of insects.)Reliability may be improved by clarity of expression (for writtenassessments), lengthening the measure,
and other informal means.However, formal psychometric analysis, called the item analysis, isconsidered the most effective way to increase reliability. This analysisconsists of computation of
item difficulties
and
item discrimination
indices, the latter index involving computation of correlations between theitems and sum of the item scores of the entire test. If items that are toodifficult, too easy, and/or have near-zero or negative discrimination arereplaced with better items, the reliability of the measure will increase.
•
R
(
t
) = 1 −
F
(
t
).
•
R
(
t
) = exp( − λ
t
). (where λ is the failure rate)
] Classical test theory
true score
and the variation of the
observed score
. Or,equivalently, one minus the ratio of the variation of the
error score
andthe variation of the
observed score
:where ρ
xx
'
is the symbol for the reliability of the observed score,
X
; ,, and are the variances on the measured, true and error scores