CHAPTER 6: QUESTIONNAIRE PILOT TESTING
6.2 Validity
Validity refers to the extent to which an instrument measures what it is supposed to measure (Bryman and Bell, 2007). A measure’s validity relies on the definitions of the variable which is used to design the measure. There are four types of validity: namely, face validity; content validity; criterion-related validity; and construct validity. These are discussed as follows (Bordens and Abbott, 2011).
6.2.1 Face validity
Face validity is the degree to which the instrument appears, on the face of it, to be an appropriate measure in obtaining the desired information from the perspective of a potential respondent. This means that questions appears to relate directly to the construct. Therefore, they should produce a valid response (Colton and Covert, 2007). Face validity is concerned with whether or not the measure "looks valid" to the respondents (Bornstein, 1996). Face validity is a subjective assessment since it depends on the judgment of experts who check the tool for grammar; suitability; and confirmation that it appears to flow logically. Therefore, it is considered to be the weakest form of validity (DeVon et al., 2006).
128 6.2.2 Content validity or Expert validity
Content validity is “the extent to which a test represents the universe of items from which it is drawn and it is especially useful when evaluating the usefulness of tests that sample a particular area of knowledge” (Salkind, 2010). Alternatively, content validity is the extent to which the indicators measure the different aspects of the concept (De Vaus, 2007). Expert validity is achieved by inviting experts, in a particular topic, to evaluate it. The measure should include adequate coverage of the subject being studied. Content validity depends on the quality of the literature and the theories which are used to build this instrument and some experts should assess the questionnaire, also, in order to determine whether or not the questionnaire measures what it should measure (Ruane, 2005; Vogt, 2007). This research tested the relationship between IC and innovations. After analysing the literature review, the study was conducted through four actors named HC, SC, OC and CC which might affect product, process and organizational innovations.
The initial questionnaire was delivered or emailed to five lecturers/senior lectures / professors in Plymouth University’s management department. They had different specializations such as human resource management; information technology; knowledge management; and banking. At the same time, it was checked by ten doctorate students specializing in business management in order to check how well they could understand the questions. Most of the feedback confirmed that the items related to their constructs and the students recommended that some items be rephrased in order to be clearer and more understandable. A member of staff recommended that the questionnaire should be translated into the Arabic language which was the respondents’ mother tongue. Therefore, the next process related to this issue.
129 6.2.3 Translation of Questionnaire
It is necessary for researchers, who apply their studies to a different language context, to translate the original questionnaire into the target language. The researcher employed back- translation as a technique to obtain a target questionnaire (Saunders et al., 2009). Back- translation means that the source questionnaire is translated into the target questionnaire (e.g. Arabic questionnaire). The final questionnaire was translated, also, into the original questionnaire (e.g. English questionnaire). Then, the researcher compared the two original questionnaires to create a final questionnaire.
The researcher sent the Arabic questionnaire by email to three professors of human resource management, marketing and banking at Business Management Department- Mansoura University- Egypt. They recommended that some words be changed so that they were clear to Egyptian banks’ managers.
6.2.4 Construct Validity
Construct validity is the extent to which items reflect the concept whereby these items are used to measure it (Howitt and Cramer, 2005). Many concepts are not measured or observed directly and, therefore, the instrument measures the constructs. Construct validity is necessary in order to check on the perceived overall validity of the measure. It is expected that a measure has high construct validity if it is built well on some theoretical construct (Clark- Carter, 2004). Colton and Covert (2007) divided construct validity into two sub-types: namely, convergent validity; and discriminant validity which are related concepts. They were assessed in the measurement model.
130
Firstly, convergent validity refers to the extent to which the correlation between the items of a construct exists strongly or, in measuring a construct, convergent validity is an association between indicators which are theoretically similar (Bergh and Ketchen, 2011). In order to obtain convergent validity in a construct, there should be high correlation coefficients (Brown, 2006; DeVon, 2007). The indicators measure the same concept. Average variance extracted (AVE) is employed to evaluate convergent validity. AVE means the overall amount of variance in the items accounted for a construct (Hair et al, 2010).
The researcher used the following formula to calculate AVE:
Where:
• λi represented the standardised factor loadings between a variable and its indicators (Factor loading for each construct),
• Var. related to variance
• ԑ was the measurement error of the indicators of construct.
In order to indicate sufficient convergent validity, the AVE should be greater than 0.5 (Dalgaard, 2008). If the researcher has convergent validity issues, this is because, within their variable, the items do not correlate well with each other; i.e., the latent factor is not explained well by its observed variables. Moreover, Composite Reliability (CR) reflects how error affects the scale. It confirms the validity of the constructs (Field, 2009). Hair et al. (1998) reported that the CR ought to be equal to or more than 0.70 to become acceptable. The researcher calculated the CR by using the following equation:
AVE=
∑λ
i2131
Secondly, discriminant validity refers to the extent to which the constructs differ from other related constructs (Tanaka, 1987; Tarling, 2009; Hair et al., 2010). Discriminant validity exists if there is no strong relationship between the constructs (Colton and Covert, 2007). Each construct should be distinct from other constructs. Therefore, high discriminant validity provides evidence that a construct is unique (Hair et al., 2010). Discriminant validity is evaluated by the square root of the AVE; this must be greater than the correlations between the constructs (Fornell and Larcker, 1981). If, for each construct, the AVE is greater than its shared variance (which is the amount of variance that a variable (construct) is able to explain in another variable) with any other construct, discriminant validity is supported. If the researcher has discriminant validity issues, this is because his/her items correlate more highly with items outside their parent factor than with the items within their parent factor; i.e. the latent factor is explained better by some other items (from a different factor), than by its own observed variables.