Abstract
Although the Youth Self-Report (YSR) has been used in many studies throughout the world, little is known about the equivalence of the factor structure of this instrument for immigrant adolescents. Measurement invariance of the three internalizing syndrome scales of the YSR was tested across four ethnic groups (native Dutch, Surinamese, Turkish, Moroccan) and over time. Results of the present study showed that the scales were invariant across all ethnic groups and over time. Together, the results indicated that the YSR can be used for developmental studies in these immigrant populations.
Notes
1 All immigrant adolescents were able to fill out the questionnaire in Dutch, because all of them were educated in Dutch.
2 A model with the items specified as being continuous showed a worse fit to the data when compared to the categorical specification. We also specified the items as having count distribution, but such a model resulted in a worse fit. The results of both models can be requested from the first author.
3 Suriname is a former Dutch colony which enabled Surinamese people to migrate to the Netherlands mainly for educational purposes starting from the 1950s. Since the 1960s, Turkish and Moroccan immigrants came to the Netherlands as labour migrants. The aforementioned groups belong to the largest immigrant groups in the Netherlands (De Valk, Citation2010)
4 Note that our analyses differed somewhat from the steps described by Van de Schoot and colleagues (Citation2012), because we analysed categorical data and therefore had to deal with thresholds instead of intercepts and residuals (see, for example, Agresti, Citation2002). Thresholds can be interpreted as the turning point where the probability of answering with the first category is no longer highest, but the probability of answering with the second category is highest. So, with three answering categories, two thresholds are estimated.
5 In the models in which the thresholds were allowed to be freely estimated across the ethnic groups, some parameters were fixed automatically by Mplus to avoid singularity of the information matrix due to empty cells in the joint distribution of the categorical variables. Furthermore, in the baseline model in the subsample too many parameters were estimated and therefore the model might not have been identified.