1. Non Positive  Definite

Jika non positive definite ditemukan pada CFA, artinya kita tidak perlu memisahkan factor/subfactor sebagai dimensi yang berbeda.

“The correlation of 1.0 between the affective and anxiety factors meant that the covariance matrix for the latent factors in the four-factor model at age 18 was not positive definite. This is further evidence that this model is overparameterized—that there is no need to split the internalizing factor into affective and anxiety subfactors (cf. Wothke, 1993).”

Krueger, Robert F.; Caspi, Avshalom; Moffitt, Terrie E.; Silva, Phil A.
Journal of Abnormal Psychology, Vol 107(2), May 1998, 216-227. http://dx.doi.org/10.1037/0021-843X.107.2.216

2. Dikotomisasi

The effect of patient race and socio-economic status on physicians’ perceptions of patients

M Van Ryn, J Burke – Social science & medicine, 2000 – Elsevier
Response options ranged from ‘not at all likely’ to ‘extremely likely’ on a five-point scale.
Due to heavily skewed response distributions for most of the single-item ordinal
measures, each was transformed into a dichotomous variable.

2. Validitas dan Reliabilitas Melalui Penilaian Rater

Interrater agreement is defined as the extent of the match between 2 or more raters’ scores assigned to performances, behaviours, or observations; whereas, interrater reliability is the extent to which scores obtained from 2 or more raters, scorers, observers, or judges are consistent (Goodwin, 2001).

Interrater agreement is based on the agreement among the judges for the degree of the rating an object should be given, interrater reliability is based on the degree that that rating by each rater can be reliably reproduced. Concordantly, “indexes of agreement are conceptually different from reliability estimates derived from either classical test theory or G theory” (Goodwin, 2001, p. 16).

LYNDA  GAIL THIESSEN. (2004). Measurement of post traumatic stress disorder: an a posteriori content validation of t h e clinician-administered ptsd scale (caps-dxsx).


A fundamental definition of validity states that “a test is valid for measuring a theoretical attribute if and only if variation in the attribute causes variation in the measurement outcomes through the response process that the test elicits” ( Borsboom & Mellen-bergh, 2007 ).

herefore, if test items are built to inherit these features as part of the explana-tion of their difficulty, the test will potentially have more construct relevant variance built into it ( Messick, 1980 ).


Comments are closed.

Set your Twitter account name in your settings to use the TwitterBar Section.