9% ( Burrai et
of 0.82>0.93 with excellent quality in al., 2012 ) and 71.2% ( De Witte et al.,
Watson et al. (2014) and 0.95 (0.80>0.96) 2011 ) for the CLES tool, and 72.8% (
with good quality in Tomietto et al. (2012) . Bergjan and Hertel, 2013 ) for the CLES + T
CLES and SECEE have demonstrated tool. Moreover, when considering only the
similar internal consistency coefficients but original versions of tools, the majority
with poor methodological quality ( De Witte reported good explained variance (from 50%
et al., 2011; Saarikoski and Leino-Kilpi, to 60%) but the methodological quality was
2002 ). poor or fair; in fact, only six studies
4.3.3. Keandalan estimated this psychometric property by
Taking reliability as the proportion of the adopting good methodology: Newton et al.
total variance in the data that is due to true (2010) reported an explained variance of
differences among learning environments as 51% in validating the CLEI; Salamonson et
well as the extent to which scores are the al. (2011) reported an explained variance of
same for repeated measurements ( Mokkink 63.3% in the CLEI-19; Saarikoski and
et al., 2010 ), only three studies ( Hosoda, Leino-Kilpi (2002) , Saarikoski et al. (2005)
2006; Gustafsson et al., 2015; Tomietto et reported 64% for the CLES instrument,
al., 2009 ) performed a test-retest evaluation while Tomietto et al. (2012) and
as reported in Table 2 , with poor or fair Papastavrou et al. (2015) reported 67.2%
methodological quality. and 67.4% respectively, in validating the
4.3.4. Measurement error CLES + T. Finally, only one study estimated
Measurement error, which is the systematic the structural validity using an excellent
and random error of a respondent score not quality methodological approach, achieving
attributed to true changes in the 58.2% of variance in validating the CLES +
construct under measurement ( Mokkink et T ( Watson et al., 2014 ). As reported in
al., 2010 ), was reported only by Gustafsson Table 2 , some authors (eg, Hosoda, 2006;
et al. (2015) in validating the CLES + T, as Papastavrou et al., 2010, 2015; Saarikoski
reported in Table 2 . and Leino-Kilpi, 2002 ) used Exploratory
4.3.5. Structural validity Factor Analysis (EFA); others (eg, Newton
The majority of studies (21 out of 26) et al., 2010; Salamonson et al., 2011 ) used
assessed the structural validity as required Principal Component Analysis (PCA) or
by the COSMIN procedures. More precisely, both (eg, Newton et al., 2010; Salamonson
the structural validity is the degree to which et al., 2011 ), whilst still others (eg, Bos et
the scores of an instrument are an adequate al., 2012; Tomietto et al., 2012; Vizcaya-
reflection of the dimensionality of Moreno et al., 2015 ) used Confirmatory
the construct to be measured ( Mokkink et Factor Analysis in addition to EFA and PCA.
al., 2010 ). struktural validity findings, when More recently, D'Souza et al. (2015) used
reported, were concordant with the Structural Equation Modeling. Moreover,
construct (dimensions) of the instrument, but Hosoda (2006) considered only students'
the methodological quality that emerged was question- naires when performing the EFA,
poor or fair in 14 of 21 studies due to whereas students and staff nurses' data were
insufficient sample size, no explanation with considered together by Dunn and Burnett
regard to the treatment of the missing items (1995) . Differently, Chuan and Barnett
and lack of precision in reporting the (2012) did not specify whether they
performed analysis. The highest explained had considered the data collected from
variance estimated by studies included in students and from educators differently.
Thus, findings regarding the structural Konvergen
validity estimations are not comparable Validity, r
given the differences in methodological Pearson
evaluation analyses performed and in the kriterium
quality of the methods used. Keabsahan
4.3.6. pengujian hipotesis Cross-
Hypotheses testing assessment as expected kultural
mean differences Keabsahan
between groups or as expected correlations a Cronbach total score
between instrument and/or range across
scores and other variables, such as the scores faktor
of other instruments ICC
( Mokkink et al., 2010 ) were estimated in SEM, SDC
eight studies out of 26 iya nih
with poor or fair methodological quality ( Perbedaan
Table 2 ). Menurut explained%,
findings, three levels of variables were metode Sebuah