RESUMO
El objetivo del estudio realizado fue evaluar las propiedades psicométricas de una medida de deseabilidad social (8 ítemes) para niños, elaborado en habla hispana (Lemos, 2005), y que se distingue por su brevedad, respaldo de la literatura científica, novedad y originalmente construido en habla hispana. El instrumento es particularmente relevante para la investigación que usa el auto-reporte como medida de recolección de datos en niños, la cual es muy frecuente en niños y adultos. Sin embargo, hay apenas algunos estudios en niños con este constructo en habla hispana, especialmente con el desarrollo de test para su medición. La validación original muestra propiedades psicométricas que posiblemente no son replicables por cuestiones metodológicas. Debido a la importancia de describir y cuantificar este constructo en población infantil, se evaluó su estructura interna (análisis factorial confirmatorio de grupo múltiple), validez de constructo (diferencias de edad y sexo) y confiabilidad (consistencia interna, KR-20 con modificación Horst) en una muestra peruana de 202 niños de nivel primario de una institución educativa pública, de educación regular. Los análisis mostraron que solo una dimensión retiene mejor la variancia de los ítemes (contrastando con la bidimensionalidad original), algunos ítemes tienen bajas correlaciones inter-ítem y cargas factoriales, la consistencia interna es baja, y que las correlaciones con la edad no siguieron el patrón esperado. Se discuten los resultados estructurales en el marco del sesgo producido por el fraseo de los ítemes y las diferencias metodológicas del estudio que se informa, aspectos que pueden influir en las propiedades psicométricas del instrumento.
The importance of the effect of social desirability on self-reported measures in children has been analyzed for several decades, with the development of instruments for children (eg, Crandall, V. C.,Crandall, V. J. & Katkovsky, 1965). This type of research is still ongoing and developing basic research (Haghighat, 2007), especially in the creation of short measures (Carifio, 1992; Haghghat, 2007) with good discrimination and masked within test battery for use in group research in children. However, research with this construct is more abundant in adults, and less common in children, especially in Latin America. May be the limitations in reading comprehension, development aspects in the recognition introspective cognitions and emotions, among other things, do not allow a deeper analysis and safety compared with adults. Also, in Spanish language, it is rare to find independent instruments that describe social desirability in children, but rather integrated into a multidimensional test (i.e., Eysenck, S. B. G. & Eysenck, H. J., 1973). One of the most important measures for social desirability speaking is developed in Hispanic context is the Children's Social Desirability Scale (EDESI -Lemos, 2005). This instrument has a particular relevance to psychological research, which calls attention to the impact of social desirability on reports that children make their own psychological attributes using questionnaires. Since EDESI can make an important psychometric contribution to quantify the social desirability in Hispanic children, ongoing evaluation of its psychometric properties have not only local value (i.e., relevant to the geographical region in which it was built) but also in the international context Hispanic. This study assess ed the replicability of results Lemos (2005), with this potential instrument of social desirability. Although some features of the study design Lemos not repeated here (e.g., distribution in different socioeconomic levels), the evidence obtained provides additional evidence in a rather partial replication framework. The sample consisted of 202 students from grade 4 to 6 (55% male and 45% female) primary level from a public educational institution of Lima, Perú, between 8 and 12 years old. For analysis, exploratory factor analysis in ULS method, parallel analysis and Scree Test for number of factors, polychoric correlations and Promax rotation (k = 4) was used. The estimate of reliability was made with the KR-20 coefficient and confidence intervals, and Horst's (1953) correction was used to reduce the impact of the dispersion of item difficulty (Merino & Charter, 2010). A one-dimensional structure as the most suitable was found to represent the construct. Internal consistency was slightly lower compared to that found in the original study, but still be considered equivalent. The results presented in this study suggest that previous findings (Lemos, 2003, 2005) are limited in their replicability, and probably the methodological choices for analyzing the internal relations of its items are the explanations for this. The instrument may require items to be scaled in three or four response choices and a slightly higher number of items if one want to improve efficiency in mass applications. Additionally, a check to verify the factor structure of the scale settings as a first step (Merino & Arndt, 2009) compared to more advanced methods (e.g., structural equations), should be conducted. According to results in this study, should be considered not just a two-dimensional model, but include a one-dimensional model that seems to compete more successfully with the original two-factor model. An important aspect is the phrasing of the items as grouping factor of the items. This problem is recognized in self-report instruments that have phrased items on direct and opposite direction to construct. The prospect of research on this topic, and with this instrument should include structural hypotheses as bifactor model and method factor as plausible explanations of relations between items. Also, assess the measurement model forget the best estimation of reliability. It can also anticipate constructing items with Hispanic multicultural relevance.