Validity and reliability of the Thai version of the ... - ResearchGate

Singapore Med J 2011; 52(2) : 105similar to that previously reported by Ciechanowskiet al. (33) The Thai ECR-R showed acceptable internalconsistency, test-retest reliability, construct andconvergent validity. The factors extracted in our studycoincided with the study of Tsagarakis et al, (15) whereseven factors initially accounted for 60% of the variance.Their study was the first investigation in a non-Anglo-Saxon language and cultural context, and it was found tobe suitable in terms of psychometric properties. Our studyyielded similar results, confirming that the translatedECR-R still possesses acceptable reliability and validitywithin a different culture. The mean figures for the ThaiECR-R were avoidance 2.89 ± 0.95 and anxiety 3.46 ±0.99, while in the original study by Tsagarakis et al, theavoidance score was 3.02 ± 0.93 and the mean scoreG-ECR-R for anxiety was 3.60 ± 1.04. (15)With regard to the model’s fit, we found both goodand poor indices. Besides a relatively small sample size,a model’s lack-of-fit may be attributed to some items thataddressed extraneous content, which do not effectivelyrepresent the latent construct for which they were written.However, the two-factor model was significantly moreappropriate than the single-factor model, while there wasno difference between models with two factors and thosewith more than two factors. Thus, it could be concludedthat the two-factor model was the most relevant to thehypothesised model. Although the model fit did not yield asatisfactory result, and thus requires further investigation,we discuss some hypotheses for the lack-of-fit here.The inappropriate values of SRMR and NFI indicatedboth simple and complex model misfits, and these maybe explained by large, standardised residuals, which ledto item sharing (i.e. items 12 and 3) or low R 2 values(e.g. items 5, 9, and 23 for the anxiety factor and items11, 13, 17 and 24 for the avoidance factor). These itemswere poorly associated with the underlying dimension andfailed to contribute to the scale. In addition, some multidimensionalitems also had an impact on the level of fit.Interestingly, our results support the recent studycarried out by Lo et al, (34) in which they found that theitems that caused the model’s lack-of-fit were similar.In Lo et al’s study, a higher-order factor analysis of amodified ECR was used. They found that the ‘discomfortwith closeness’ factor, which should represent anavoidance orientation, had a double-loading. This factorcomprised the items corresponding to our problematicitems, i.e. items 2, 3, 11, 17 and 24. In addition, the itemaddressed as extraneous content in Fairchild and Finney’sstudy (17) was found to play the same role in our study, i.e.item 12 (“I’m afraid that once a romantic partner gets toknow me, he or she won’t like who I really am”). Finally,a new set of 28-item questionnaire, as recommended bymodification indices, was re-analysed. Some items wereremoved from the new set to ease administration. Thesewere items 24, 17, 13 and 11 on the avoidance factor anditems 23, 12, 9 and 5 on the anxiety factor. The resultsturned out an adequate model fit, as follows: CFI = 0.99,NFI = 0.86, RMSEA = 0.021, SRMR = 0.07.In summary, the Thai ECR-R demonstrates goodinternal consistency and fair-to-good test-retest reliability.The lack of fitness of the model suggests that some itemson the scale were inappropriate in the representation ofthe constructs of anxiety and avoidance. Deleting someitems provided a better construct validity. A stand-aloneexamination of the 28-item ECR-R (or less items) shouldbe conducted, particularly with a more general sample ofthe population.ACKNOWLEDGEMENTSThis research was funded by the Faculty of Medicine,Chiang Mai University, Thailand. The authors wish tothank Assistant Prof Dr Ruetinan Samuttai, Faculty ofEducation, Chiang Mai University and Dr NumchaiSuppareakchaisakul, Srinakharinwirot University,Bangkok for their suggestions regarding the confirmatoryfactor analysis and Lisrel programme.REFERENCES1. Bowlby J. Attachment and loss. Volume 2. Separation: Anxietyand anger. New York: Basic Books, 1973.2. Bowlby J. Attachment and loss. Volume 3. Loss: Sadness anddepression. New York: Basic Books, 1980.3. Bowlby J. Attachment and loss. Volume 1. Attachment. 2ndedition. New York: Basic Books, 1982.4. Johnson LN, Ketring SA, Rohacs J, Brewer AL. Attachment andthe therapeutic alliance in family therapy. Am J Fam Ther 2006:34:205-18.5. Naaman S, Pappas JD, Makinen J, Zuccarini D, Johnson-DouglasS. Treating attachment injured couples with emotionally focusedtherapy: a case study approach. Psychiatry 2005; 68:55-77.6. Parish M, Eagle MN. Attachment to the therapist. PsychoanalPsychol 2003, 20:271-286.7. Bowlby J. The making and breaking of affectional bonds. London:Tavistock, 1979.8. Ainsworth MDS, Blehar MC, Waters E, Wall S: Patterns ofAttachment: A Psychological Study of the Strange Situation. NewJersey: Hilldales, 1978.9. Bartholomew K, Horowitz LM. Attachment styles among youngadults: a test of a four-category model. J Pers Soc Psychol 1991;61:226-244.10. Bartholomew K, Griffin DW. The metaphysics of measurement:the case of adult attachment. In: Bartholomew, K and Perlman,D (Eds). Advances in Personal Relationships. London: JessicaKingsley, 1994;5:17-52.11. Brennan KA, Clark CL, Shaver PR. Self-report measures ofadult romantic attachment. In: Attachment Theory and CloseRelationships J.A Simpson JA, Rholes WS, eds. New York:Guilford, 1998: 46-76.12. Fraley RC, Waller NG, Brennan KA. An item response theory