Skip navigation

Category Archives: Construct Development

Article by Petter, Stacie; Straub, Detmar, and; Rai, Arun (2007) in MIS Quarterly, 31:4.

 

In the previous entries, I have posted my notes on two articles pertaining to construct development, i.e. Law, et. al. ( 1998 ) and Henri ( 2008 ). This entry is the third one. For better understanding, I would like to advise readers to read this article first, followed by Law et. al. ( 1998 ) and lastly Henri ( 2008 ).

 

If I was given a chance to teach research methodology for design and social science, I would make it compulsory to all my students to read this article together with the one by Law et. al. Both articles described the fundamental knowledge and the essential basics in developing or conceptualizing a construct.

 

The first most important point of discussion in this paper is about distinguishing two types of construct, i.e. formative and reflective. I have summarized the different characteristics between them in the following table for easy understanding.

Formative vs Reflective

Formative vs Reflective

 

 

 

 

 

The next point is about the effect of misspecification of formative construct on type I and type II errors. It is worth to note here that misspecification of measurement model would lead to measurement error, which in turn affect badly the structural model. Thus, it would increase the potential for both type I and type II errors, which no doubt would make the research findings equivocal.

 

The authors used simulation technique (as extension of earlier simulation done by Jarvis et. al. & MacKenzie et. al.), where the result had revealed the combination of characteristics of formative constructs that would possibly cause the type I and type II errors. Those combinations of characteristics are as follow.

 

Combination of characteristics that may cause type I error:

  1. Formative construct is endogenous
  2. Structural path emanates from formative construct
  3. Sample size is high (i.e., 500)
  4. Moderate to high correlation among formative measures/items (i.e., 0.4 or higher inter-item correlations)
  5. Can occur regardless of whether the model is specified correctly (i.e. formative) or incorrectly (i.e., reflective).

 

Combination of characteristics that may cause type II error:

  1. Formative construct is endogenous
  2. Structural path leads to formative construct
  3. Sample size is low (i.e., 250)
  4. Moderate to high correlation among formative measures (i.e., 0.4 or higher inter-item correlations)

- OR -

  1. Formative construct is endogenous
  2. Structural path leads to formative construct
  3. Sample size is high (i.e., 500)
  4. High correlation among formative measures (i.e., 0.7 or higher inter-item correlations)

 

Another important contribution of this paper is the recommendation on steps (decision rules) to be taken in identifying formative constructs. The authors had put the steps in a well organized table, so that readers can understand it easily. Here is the table:

Identify Formative Construct

Identify Formative Construct

Article by Law, Kenneth S., Wong, Chi-Sum, and Mobley, William H. ( 1998 ) in Academy of Management Review, 23:4.

 

This paper classified multidimensional constructs into 3 different types, i.e. latent model, profile model and aggregate model. The basis used for the classifications were: (i) the relations between construct and its dimensions, (ii) the interpretation of the nature of the construct, and (iii) the definition of true variances of the construct.

 

Researchers need to consider the appropriate type of the construct as different type carries different interpretation of nature, different definition of true variance, and different theoretical meanings. The authors argued that proper conceptualization of constructs with regards to this classification is very important for three main reasons:

  1. Definition of the Research Question – “without correct specifications of the relations between multidimensional constructs and their dimensions (the main basis of argument of this paper), one would set up research hypotheses at the construct level, conduct analyses at the dimensional level but make conclusions at the construct level”.
  2. Theoretical Parsimony – “only when the interrelations between a multidimensional construct and its dimensions are specified can we derive overall and parsimonious conclusions about the role of the multidimensional construct in its nomological network”.
  3. Relations with Other Constructs – Different specifications of the relations between the multidimensional construct and its dimensions were proven resulting differences in parameter estimates. Therefore, the specifications must be done properly as incorrect specifications during the construct conceptualization will definitely lead to wrong conclusions.

 

I prepare my note in diagram form for easy understanding, and I put it here for everybody to understand well about the three models of multidimensional construct:

Multidimensional Construct

Multidimensional Construct

 

Follow

Get every new post delivered to your Inbox.

Join 28 other followers