Skip to content
This repository was archived by the owner on Jun 1, 2023. It is now read-only.

Research Model

Jump to bottom
jialincheoh edited this page Jun 20, 2022 · 81 revisions

Updated Research Model 5/13/2022

Research Questions

What are the joint effects of performance and solution transparency on the innovation outcome?

Existing Theories

  • Solution transparency leads to convergence (Bodreau & Lakhani, 2014)
  • The groups with performance transparency will have reference points that motivate them to jump and innovate if they are below the best. As a result, they will take more risks and explore more (Brunswicker & Prietula, 2017). Social aspiration will make people with performance information take more risks (March, 1988).

Theory for Transparencies

Level of Analysis

Level of analysis is done at the individual level.

Dependent Variables

  1. Innovation Performance
  • novelty-phaseX -> novelty in phase X
  • user-requirement-phaseX -> user requirement score in phase X
  • infovis-phaseX -> information visualization score in phase X
  • tech-phaseX -> technology score in phase X
  • improve-novel-XY -> improvement in novelty score from phase X to phase Y
  • improve-ur-XY -> improvement in user requirement score from phase X to phase Y
  • improve-vis-XY -> improvement in information visualization score from phase X to phase Y
  • improve-tech-XY -> improvement in technology score from phase X to phase Y
  • improve-total-XY -> improvement in total score from phase X to phase Y
  1. Effort
  • add-loc-XY -> lines of codes added from phase X to phase Y
  • delete-loc-XY -> lines of codes deleted from phase X to phase Y
  • change-files-XY -> number of files changes from phase X to phase Y
  • improve-effort-XY -> improvement score calculated with 1*[number of lines added] - 0.5*[number of lines deleted]
  • sum-add-loc -> sum of lines of codes added across all phases (phase12 + phase23 + phase34 + phase45)
  1. Exploration

3.1 Variety and Uniqueness

  • Concept: Distance of search (Billinger et al. 2014; Brunswicker et. al. 2019; ). Cross stage similarity metrics on an individual level (Variable Name: individual_exploration (IND_EXP)). In addition, we will have template similarity (Variable Name: template_similarity (TEMP_SIM)).
  • Concept: Relational Novelty (Kyriakou et al. 2021 ). Dissimilarity with others on a prior phase (Variable Name: stageX_bogotaX_high_similarity (stageX_bogotaX_high_similarity))
  • Concept: Variety/Scope (Schiling et. al. 2011). Count of functions for each individual at each group and at each phase [ non-repetitive ] (Variable Name: unique functions for each individual without repetition (unique))
  • Concept: Rareness measure (Schiling et. al. 2011; Uzzi et al. 2013). Rareness of functions (top 5%, 10%, 15%, 20% most unique functions pooled globally) (Variable Name: len_unique_bogotaX_phaseX)

Covariates

  1. Attention
  • phaseX-clicks-count -> clicks done by a person in a specific phase
  • sum-clicks-count -> sum of clicks done by a person across all phases
  • project-clicks-hacker -> number of clicks obtained by a particular hacker for his code for each person, each phase, each group
  • project-clicks-count -> number of project clicks done by a particular person for each phase and each group.
  • score-clicks-count -> number of score clicks done by a particular person for each phase and each group.
  • cum-bogotaX-phaseX -> sum of the total scores of the projects clicked on by a particular user in a phase and in a group.
  • unique_apps -> number of unique apps clicked on by a particular user in a phase and in a group ( removing repetition ).



[ Archive ] First Research Models

Research Questions - What are the joint effects of performance and solution transparency on the innovation outcome?

Dependent Variables

  1. Innovation Performance
  • Only the improve-novelty-15
  • Compound Measure of improve-novelty-15, improve-tech-15, improve-ur-15, improve-vis-15 ( PCA / EFA )
  • Compound Measure of improve-novelty-15, improve-ur-15 ( Ammabile - novel and useful )
  1. Effort
  • added-loc-15 ( priority )
  • effort-loc-15
  • delete-loc-15
  1. Exploration
  • New functions used, which are not used in the prior round.
  • Visualization Components Used ( Yes / No )
  1. Error Rate
  • Diff between stage 1 and stage 5 errors

Treatment variables are made up of 2 constructs - performance and solution transparency

  • 4 treatments ( No Transparency, Performance Transparency, Solution Transparency, Full Transparency )
  • 2 treatments recoded

Hypotheses

H1 - Full transparency has the highest innovation performance compared to all the other treatment groups. H2 - Full transparency has the highest risk-taking behaviors compared to all the other treatment groups. H3 - Solution transparency H4 - The interactions of performance and solution transparency strengthen the effect of transparency independently.

Covariates

  • Motivations
  • Aspirations ( measured with some kind of click measure + performance difference )
  • Skills
  • Prior Experience
  • Attention ( measured with some kind of click measure )

Model Test Run with Planned Contrast Technique

Now, the ANOVA we want to do is to see whether the treatment has any impact on the performance such as 'improve-tech-15'. We also want to aspiration into account … technically making this an ANCOVA. We run the below commands to get the output.

library(car)
aov1 = aov( improve-tech-15 ~ treatment + aspiration, df)
Anova(aov1 ,type="III")

This yields the following result.

So we can see that treatment does influence improve-tech-15. But how exactly? Planned contrasts can help us find out more about this. However, planned contrasts require that we really understand our independent variable of interest.

Revised Research Models

Clone this wiki locally