A generalization of this approach to analyzing response patterns is item response theory (IRT) (56; 57), occasionally also called latent trait analysis (58). This is based on a measurement model developed by Georg Rasch for assessing abilities that is now frequently applied to health measurements. The “latent trait” refers to a theoretical continuum that the test items are designed to measure. Using physical function as an example, the continuum running from fitness to severe disability (the latent trait) is plotted along the horizontal axis of a graph. Items that measure different severities of disability are spaced along this continuum; items such as “I can run a mile” would be near the left, and “I am confined to a wheel chair” near the right. The Rasch model is based on several postulates:
(i) Scores on a measure depend on the ability of the person and the level of disability implied by the items.
(ii) A good scale should have items that range in difficulty, and with a rank order of difficulty that does not vary from respondent to respondent.
(iii) Good measurement requires that a person’s ability be accurately reflected regardless
of the scale used.
(iv) Where a measurement contains several scales, such as physical, mental, and social, these should each be unidimensional, because if questions on different topics are combined, the resulting score cannot be clearly interpreted (unless the score is extreme).
A generalization of this approach to analyzing response patterns is item response theory (IRT) (56; 57), occasionally also called latent trait analysis (58). This is based on a measurement model developed by Georg Rasch for assessing abilities that is now frequently applied to health measurements. The “latent trait” refers to a theoretical continuum that the test items are designed to measure. Using physical function as an example, the continuum running from fitness to severe disability (the latent trait) is plotted along the horizontal axis of a graph. Items that measure different severities of disability are spaced along this continuum; items such as “I can run a mile” would be near the left, and “I am confined to a wheel chair” near the right. The Rasch model is based on several postulates:(i) Scores on a measure depend on the ability of the person and the level of disability implied by the items.(ii) A good scale should have items that range in difficulty, and with a rank order of difficulty that does not vary from respondent to respondent.(iii) Good measurement requires that a person’s ability be accurately reflected regardlessof the scale used.(iv) Where a measurement contains several scales, such as physical, mental, and social, these should each be unidimensional, because if questions on different topics are combined, the resulting score cannot be clearly interpreted (unless the score is extreme).
การแปล กรุณารอสักครู่..
