1. Moshe Banai, PhD Moshe.banai@baruch.cuny.edu Source: The Community-Based Intervention Research Group (C-BIRG), of the Robert Stempel College of Public Health and Social Work at Florida International University

2. Translations Should we translate questionnaires? The Indigenous Psychology Movement Emic & Etic Conceptualizations Emic constructs ‐limited to a single culture (Guilt vs. Shame) Etic constructs ‐exist in identical or near identical from across a range of cultures (The norm of reciprocity exists in almost all cultures) Extreme versions of the indigenous psychology movement embody an error.

3. Translations Can we Translate Questionnaires? Lack of Semantic Equivalence across Languages Lack of Conceptual Equivalence Across Cultures Lack of Normative Equivalence Across Societies

4. Characteristics of Useful Measures Semantic Equivalence (Involves the choice of terms and sentence structures that ensure that the meaning of the source language statement is preserved in the translation) Conceptual Equivalence (refers to the degree to which a concept, independent of the words used to operationalize it, exists in the same form in the source and target cultures) Normative Equivalence (refers to the degree to which the researcher has dealt successfully with the problems created by differences in societal rules)

5. Lack of Normative Equivalence Across Societies Willingness to discuss certain topics Manner in which ideas are expressed Treatment of strangers

6. Background and demographic data problem likelihood Semantic: LOW: with care, its is possible to pick target language words that yield correct facts. Conceptual: LOW: Concepts or Constructs are not normally a part of demographic information. Normative: HIGH: Societies vary in willingness to share various types of personal information with strangers.

7. Behavioral reports problem likelihood Semantic: LOW: It is generally quite easy to phrase questions in ways that clearly designate the behaviors of interest. Conceptual: LOW: Concepts or constructs are not normally a part of behavioral reports. Normative: HIGH: Societies vary widely in norms regarding what behaviors should be discussed with strangers.

8. Attitudes and Opinions Problem Likelihood Semantic: HIGH: Attempts to find words that capture abstract ideas involved in measuring attitudes require extreme care. Conceptual: HIGH: Important questions have been raised about the universality of many concepts of interest to researchers. Normative: HIGH: Societies vary widely in attitudes and questions that can properly be discussed with strangers.

9. Knowledge Problem Likelihood Semantic: MEDIUM: Selecting wording that taps the same knowledge is difficult but not impossible. Conceptual: LOW: Concepts and constructs are not normally critical parts of measurement of knowledge. Normative: LOW: Questions are rarely culturally sensitive, though the tendency to become who‐involved in “testing” process may vary from culture to culture.

10. Intentions, Expectations and Aspirations Problem Likelihood Semantic: MEDIUM: Finding equivalent words for many intentions, expectations and aspirations is simple, but complex ones may be hard to translate. Conceptual: LOW: Questions about intentions, expectations and aspirations deal mostly with concrete items not abstract concepts. Normative: MEDIUM: Societies vary as to what intentions, expectations and aspirations may be discussed with strangers.

11. Summary of Problems Likelihood Demographic BehaviorAttitudes Knowledge Intentions Semanticlow highmedium Conceptuallow highlow Normativehigh lowmedium

12. Characteristics of Useful Measures The translated version of the questionnaire MUST satisfy two sets of requirements. 1. Must meet the basic standards set for all measures, translated or not (that is Validity, Reliability, Legal and Cost) 2. Must meet the requirements for equivalence relative to the source language measure.

13. Characteristics of Useful Measures Basic Requirements Reliability, degree to which a measure of a construct is free from random error. Parallel Forms Reliability Test‐Retest Reliability Internal Consistency Reliability Reliability of Translated Measures Far Eastern cultures are (1) more likely to select a more‐or‐less neutral midpoint from among a set of response alternatives; (2) less likely to give negative responses than are their Western counterparts.

14. Characteristics of Useful Measures Basic Requirements Validity, is the degree to which a measure is free from systematic bias. Predictive validity Content Validity Construct Validity

15. Characteristics of Useful Measures Basic Requirements Utility Statistical Significance Practical Utility Economic Utility Legality, Cultural acceptability

16. Characteristics of Useful Measures Basic Requirements Equivalence Semantic Equivalence(Involves the choice of terms and sentence structures that ensure that the meaning of the source language statement is preserved in the translation) Conceptual Equivalence (refers to the degree to which a concept, independent of the words used to operationalize it, exists in the same form in the source and target cultures. Normative Equivalence(refers to the degree to which the researcher has dealt successfully with the problems created by differences in societal rules), dealing with: (a) the openness with which particular topics are discussed; (b) The manners in which ideas are expressed; (c) The way in which strangers, particularly strangers asking questions, are treated.

17. Solving Semantic Problems Solving Semantic Problems when Translating an Existing Instrument Methods researchers use to prepare target language versions of existing instruments are: Direct Translation (a)Simple direct translation (b)Modified direct translation Translation/back‐translation The ultimate test—performance criterion The parallel blind technique The random probe technique De-centering

18. Solving Semantic Problems Solving Semantic Problems when Translating an Existing Instrument They are evaluated by : Informativeness Source Language Transparency Security Practicality

19. Solving Semantic Problems

20. Solving Semantic Problems when Translating an Existing Instrument Translation/back‐translation A bilingual individual translates the source language instrument into the target language. A second bilingual individual with no knowledge of the wording of the original source language document translates this draft target language rendering into source language. The original and back‐translated source language versions are compared. If substantial differences exist between the two source language documents, another target language draft is prepared containing modifications designed to eliminate the discrepancies.

21. Solving Semantic Problems Solving Semantic Problems when Translating an Existing Instrument Translation/back‐translation Factors that can lead to similarity of original and back translated source language versions, when in fact it isn’t so: Same set of conventions for handling material that is in fact not equivalent. Back‐translation may be able to guess the intent of a poorly translated item. The draft target language version may contain elements of the source language grammatical structure that make it possible for bilingual individual to guess the source wording. Translators knew that their work was going to be subjected to back translation.

22. Solving Semantic Problems Brislin (1980) encapsulates his advice in twelve rules: Use short, simple sentences of less than sixteen words. Employ the active rather than passive voice. Repeat nouns rather than using pronouns. Avoid metaphor and colloquialisms. Avoid the subjunctive mood, e.g., verb forms with “could” or “would”. Add sentences which provide context for key ideas. Reword key phrases to provide redundancy. Avoid adverbs and prepositions telling “where” or “when” (e.g. frequent, beyond, upper). Avoid possessive forms when possible. Use specific rather than general terms. Avoid words indicating vagueness regarding some event or thing( e.g., probably and frequently). Use wording familiar to the translators where possible. Avoid sentences with two different verbs if the verbs suggest two different actions

23. Solving Semantic Problems Source Language Centering It contains of the following steps: A source language instrument is developed Instrument is translated into the target language Then a different person translates this draft instrument back into the source language Versions are compared Changes designed to eliminate the discrepancies are made in the wording of the source language instrument, the target language instrument, or both Process is repeated using other translators

24. Solving Semantic Problems Multicultural Team Approaches “Triandis Procedure”(creating parallel operationalizations involves the use of the target and source languages at the same time, rather than creating the source language version first and then deriving the target language version from it).

25. Solving Conceptual Problems Logical Tests of Conceptual Equivalence Consideration of: The constitutive definitions of the concept of interest The theory which explains it The nature of any differences between the source and target cultures

26. Solving Conceptual Problems Empirical Tests of Conceptual Equivalence Tests of Similarity of factorial Structures Exploratory Factor Analysis Confirmatory Factor Analysis Tests of the Similarity of Positions in Nomological Nets Tests of the Similarity of Scores Item Response Theory

27. Solving Conceptual Problems Empirical Tests of Conceptual Equivalence Exploratory Factor Analysis The use of exploratory factor analysis (EFA) to test for conceptual equivalence requires the comparison of factorial structures responses of comparable samples from the source and target cultures. The scales are measuring the same thing if: The numbers of factors extracted from the two versions are the same The same items load on each factor The same proportions of the total variance are accounted for by each factor The intercorrelations among the factors extracted do not differ significantly

28. Solving Conceptual Problems Empirical Tests of Conceptual Equivalence Confirmatory Factor Analysis CFA can be used to determine if the same factorial structure is present in a translated version of the scale Scales conceptually equivalent if: Original scale of the model fits the data better than the null model (i.e., one in which all items are totally unrelated) Original scale fits the data better than the one‐factor model (i.e., all items load on one single factor) when the original scale is a multidimensional one The general fit of the original scale on the target language version is within the conventional acceptance limits

29. Solving Conceptual Problems Empirical Tests of Conceptual Equivalence Tests of the Similarity of Scores Forms of score similarity: Metric equivalence exists when the properties of two versions, such as item factor loadings, are the same or similar. Scalar Equivalence means scores are directly comparable (i.e., 3.8 on, say, the Hindi version of scale is equal to 3.8 on Korean version). Item equivalence exists when individual item characteristics curves are similar across the two cultures.

30. Solving Conceptual Problems Empirical Tests of Conceptual Equivalence Item Response Theory IRT specifies a relationship between the subjects observable scores on a scale and the unobservable traits assumed to underlie the scale. The most important output of an IRT is the item characteristic curve (ICC), a plot of the relationship between responses to an item and latent construct hypothesized to underlie them. In IRT, an item is said to have differential item functioning (DIF) when the ICCs for the same items on the two versions of the scale differ by more than sampling error.

31. Solving Conceptual Problems Empirical Tests of Conceptual Equivalence Differential Item Function (DIF )testing methods: The parameter equating method test the differences between the two cultures item parameters once they have been transformed to the same metric. The Mantel‐Haenszel method tests whether the actual frequencies of specific responses for the cultural group across various trait levels differ significantly from the frequencies that would be expected if there were no differences in the odds of a specific response for the two cultural groups. The Model Comparison method is analogous to the test of factorial invariance in a confirmatory factor analysis.