Analyzing Routine Structures in Open Source Software Development Digital Traces & Qualitative Inquiry Aron Lindberg Case Western Reserve University.

Analyzing Routine Structures in Open Source Software Development Digital Traces & Qualitative Inquiry

Aron LindbergCase Western Reserve University

Research Problems

• How do OSS projects match the complexity of problems with requisite complexity of routines?

• How do OSS projects differ in terms of their performative, routinized methodologies?

Empirical Questions

1. What are the different types of routine structures in OSS projects?

2. What factors shape different types of routine structures in OSS projects?

Digital Trace Data

• Extracted from Github (https://developer.github.com/v3/issues/events/) using a 3rd party SQL database (http://www.ghtorrent.org/) and public API (http://octokit.github.io/octokit.rb/)

• Transformed using R• Analyzed using R package TraMineR

http://www.orgdna.net/traminer/

Data represented as “thin” traces (left) and “thick” text (right)

RQ #1: What are the different types of routine structures in OSS projects?

Analysis #1: Cluster Analysis

1. Run Optimal Matching (OM) algorithm on a single project

2. Identify multiple cluster solutions3. Evaluate and choose cluster solution based

on fit statistics4. Characterize each cluster quantitatively5. Qualitative inquiry into routines as text

Analysis #1: OM algorithm

Routine 1: open_issue, edit, mergeRoutine 2: open_issue, edit, close

Routine 3: comment, comment, mergeRoutine 4: comment, comment, comment

Analysis #1: Choosing a Cluster Solution

2 3 4 5 6 7 8 9 100

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

PBCHGASW

Number of Clusters

Exp

lan

ator

y P

ower

Analysis #1: Descriptive Cluster Statistics

Analysis #1: Qualitative Inquiry

RQ #2: What factors shape different types of routine structures in OSS projects?

Analysis #2: Variation

1. Run optimal matching algorithm on multiple projects

2. Plot distance matrices as heatmaps & calculate average heterogeneity of routines within each project

3. Conduct qualitative inquiry into context using interviews & archival data

Analysis #2: Heat Maps

Red = more homogenous, yellow = more heterogeneous

0.41 (0.25)

0.56 (0.16)

0.62 (0.18)

0.48 (0.25)

Analysis #2: Qualitative Inquiry

0.41 (0.25)

0.56 (0.16)

0.62 (0.18)

0.48 (0.25)

High ModularizationLow Modularization

Substantive Rationality

ProceduralRationality