The Teamwork Process Antecedents (TPA) questionnaire: developing and validating a comprehensive measure for assessing antecedents of teamwork process quality

Abstract

Context

Most models of teamwork describe team behavior and effectiveness using an Input-Process-Output approach. In software engineering, the use of such models has focused on understanding and operationalizing the Process-Output components while less research effort has been applied to define and measure the Input-Process component.

Objective

To develop and validate a measure of teamwork process antecedents (inputs) that addresses specific characteristics of software teams in industrial practice.

Method

First, we reviewed the group work literature, identified and integrated previously described antecedents of work group process, and developed a measure to tap those antecedents. This measure is operationalized in the Teamwork Process Antecedents (TPA) questionnaire, which we then validated with 375 Brazilian software engineers from 100 companies, using exploratory and confirmatory factor analysis.

Results

We created a survey to operationalize two multidimensional antecedents of teamwork process, Team Structure and Team Composition, based on well-established models from the literature on work teams. We tailored the response items to the software engineering context to increase construct face validity. We reached a parsimonious set of five dimensions for Team Composition (16 response items) and four dimensions for Team Structure (11 response items). Our results show that our measure of TPA has excellent internal reliability and convergent and discriminant validity.

Conclusions

We created a novel measure of antecedents of teamwork process tailored to software teams, that captures the perception of team members about the adequacy of team composition and structure to achieve team goals. Further, we present the development of the TPA measure in the form of a guideline that may be used in the construction of other measurement instruments in empirical software engineering research. We believe both results are important contributions of this work.

Appendices

Appendix 1

Table presents in the first three columns the set of constructs obtained by Pereira et al. (2017) and Marsicano et al. (2017) (first column), Gladstein (1984) (second column) and the ad-hoc literature review (third column). Finally, the last column presents the constructs we chose to be used in the research questionnaire. The decision was based on our experience and knowledge supported by the relevance of each construct in the literature reviewed.

Table 21 Comparing and decision related to the set of constructs to be used

Appendix 2

Table 22 List of Response Items, per 1st Level Constructs, of TPA

Appendix 3

Tables to use in Confirmatory Factorial Analysis (CFA).

Table 23 Description of the Quality Indices Classifications (Marôco 2010)

Table 24 Quality Indices, Reference Values, Description and Classification

Table 25 Indices and Conditions to verify Internal Reliability and Construct Validity

Appendix 4

Observing Table 26, it is possible to notice that all latent factors of the 1st order have a high level of significance (> 0.50) when related to a 2nd order latent factor, except for the factor of ‘experience in the organization’. Indicating the possibility of having a multifaceted, operationalized, 2nd order factor, from the 1st order factors. Table 27 presents the relation of the 1st order factors with two 2nd order factors, suggesting the existence of two groupings. In this scenario, all factors have substantive factor loads (> 0.50), except for the factor of ‘experience in the organization’. In this second context, the distribution of factorial loads between the two groups is more parsimonious when compared to the data presented in Table 26. An indication that the existence of two latent factors of 2nd order may be more adequate than just one

Observing Table 28, three main points can be noted. The first one refers to the maintenance of the grouping of 1st order factors, interpersonal skills, role and goal clarity, formal leadership, work experience, management skills and rules of behavior, with a 2nd order factor, as was reported in Table 27. This reinforces the evidence of grouping these factors into a higher-order factor. Second, 1st order factors that relate to Factor 2, of 2nd order, have unbalanced factorial loads (with a difference greater than 0.40 between them); in addition, the factor of experience in the organization has a load lower than 0.30, and can be considered statistically independent, not contributing to the factor analysis (Hair et al. 2009). Third, Factor 3 (2nd Order) is related to only one 1st order factor, therefore not justifying the possible existence of a higher-order factor. In view of these points, the possibility of having 3 or more higher-order factors related to the 9 first-order factors identified in the EFA is ruled out

Table 26 Factors Matrix with One 2nd Order Factor ^a

Table 27 Rotated Factors Matrix with Two 2nd Order Factors ^a

Table 28 Factors Matrix with Three 2nd Order Factors ^a

Appendix 5

Table 29 presents the Rho (ρ) of reliability and convergence (Rho_cv) of each factor, as well as the factorial loads of each of the items that make up the final measurement scale. Table 29 also shows the mapping of each factor with the aspects of the team (composition and structure).

Table 29 Reliability and Convergent Validity of Factors - Model 1

In Table 30, the values that show high significance of all relations between factors and items are shown, where the values of C.R. > 1.96 and p value values are all below 0.001 (***).

Table 30 Reliability and Validity of Factors - The Model of Teamwork Process Antecedents (TPA) Questionnaire