4.4 Research Methodology
4.4.2.4 Variables and Measures
In line with the hypotheses, the dependent variables of interest were requirements size,
requirements quantity, and requirements completeness (depth and breadth). Requirements
size was measured as the number of words used to construct the requirements message.
Requirements quantity was the total number of requirements provided by a participant
(e.g., Browne and Rogich, 2001). In line with Pitts and Browne (2007), requirements
completeness was comprised of the measures of depth and breadth of requirements.
Breadth was measured as the number of distinct requirements categories (from the
Browne and Rogich requirements taxonomy shown below) provided by subjects. Depth
was measured as the number of requirements within each category (from the taxonomy)
provided by subjects. The taxonomy has been used in many studies to code requirements
data (e.g., Browne and Rogich (2001), Pitts and Browne (2007)). For the purpose of
analysis involving requirements quantity and completeness, a coding procedure (similar
to Browne and Rogich (2001)) for the textual requirements data was used (described
later).
The taxonomy (Pitts and Browne, 2007) is shown in Table 24.
Table 24. Requirements taxonomy (Pitts and Browne, 2007) GENERIC REQUIREMENTS CATEGORIES
(Pitts and Browne, 2007, p.101)
DESCRIPTION (Pitts and Browne, 2007, p.101) GOAL
Goal state specification Identifying the particular goal state to be achieved
Goal specification Comparing existing and desired states
Difficulties and constraints Identifying factors inhibiting goal achievement Ultimate values and preferences Stating the final ends served by a solution Difficulties and constraints Identifying factors inhibiting goal achievement Ultimate values and preferences Stating the final ends served by a solution Means and strategies Specifying how a solution might be achieved Causal diagnosis Identifying the causes of the problematic state Knowledge specification Stating facts and beliefs pertinent to the problem
128
Existing support environment Existing technological environment to support new system
Stakeholders Organization units, customers, suppliers,
competitors
PROCESS
Process description Steps or tasks designed to produce a product Process knowledge specification Facts, rules, beliefs, decisions required to perform
process
Difficulties and constraints Facts that may prohibit process completion Roles and responsibilities Individuals/departments charged with performing
process
TASK
Task description Sequence of actions required to complete a task
Task knowledge specification Facts, rules, beliefs, decisions required to perform task
Performance criteria Statement that link outcome with
condition/constraints
Roles and responsibilities Individuals/departments charged with performing task
Justification Explanation of specific action to be/not to be taken
INFORMATION
Displayed information Data to be presented to users; paper/electronic format
Interface design Language and format used for displayed
information
Inputs Data that must be entered in to system
Stored information Data saved by the system
Objects and events Physical entities and occurrences relevant to the system
Relationship among object/event How one object/event is associated with another object/event
Data attributes Characteristics of objects and events
Validation criteria Rules that govern the validity of data
Computations Information created by the system
Some control variables that were expected to be present in the context of
requirements generation in OSS development were also measured. The experience and
knowledge of software project team members and users can influence the outcomes of the
requirements capture stage of a software project (Chatzoglou and Macaulay, 1997).
Technical knowledge (e.g., knowledge about commonly occurring functional
requirements in the context of information systems development, general knowledge
129
important knowledge sources during requirements determination for systems
development (Vitalari, 1985). Thus, one’s technical background (technical career and
technical education) can potentially influence requirements generation in OSS
development and were included as covariates in the analysis. Questionnaire items for
measuring for the extent to which the participants’ career and education were technical
were obtained from Enns et al. (2003).
Web-based requirements generation is different from the traditional face-to-face
setting of requirements elicitation, as web-based requirements generation is mediated by
the internet. Hence experience with the relevant web-based applications could be a
potential confounding variable. One such variable is the crowdsourcing experience of
participants, which is the number of crowdsourcing tasks (hosted on some crowdsourcing
platform) that participants had completed in the past. For the crowdflower platform, this
would be the total number of crowdflower tasks (tasks hosted on that platform) that a
participant has completed. This measure was obtained for each participant from their
Crowdflower profile. Crowdsourcing platforms, such as Crowdflower, are web-based
platforms/applications that support crowdsourcing tasks in a variety of domains (Doan et
al., 2011). Example categories of such crowdsourcing tasks are reviewing (e.g., books,
movies), tagging (e.g., webpages) and constructing/sharing textual knowledge (e.g.,
answering questions) (Doan et al., 2011). Participants who have successfully completed
many such crowdsourcing tasks could potentially gain knowledge of many different
application domains. They may become adept at common web-based activities such as
130
important role during requirements generation in OSS development. Hence,
crowdsourcing experience of participants was included as a covariate.
The prior experience of participants with medical software (e.g., personal health
record software) could also be a potential source of knowledge for them during
requirements generation; hence, it was also included as a covariate. Participants were
asked to indicate their usage experience with electronic medical record software systems,
personal health record software systems, and other health related software systems on a
Likert scale and the scores were aggregated and included in the analysis as the control
variable, experience with medical software.
4.5 Data Analysis and Results