Introduction to NVivo for Qualitative Data Analysis

Please see the full course outline, including diagrams and demos here

NVivo is software programme for qualitative data analysis. It is a powerful platform that supports text, multimedia, pictures, PDFs, open-ended surveys from Excel and Survey Monkey, reference libraries, webpages, social media data from Facebook, Twitter, LinkedIn, and YouTube, notes from Evernote and OneNote, and emails from Outlook. NVivo supports a range of inductive and deductive methods to qualitative analysis such as thematic and content analysis, within and cross-case analysis, discourse, conversational and narrative analysis, grounded theory, analytical induction, and qualitative research synthesis. The objective of this course is to provide participants with knowledge and skills to use the basic and advanced features of NVivo in their own research. The course content is spread over four modules and includes to set up a project and organise data, work with multimedia, manage a literature review, autocode and code data inductively, generate hypotheses, seek patterns and discover relationships, and present qualitative findings. Details of the four modules is presented below.

Module 1 Data Management

The course opens with notions of qualitative research designs and their application in a NVivo project.  In turn, we review how data can be organised in comparative and non-comparative designs, coding approaches developed, and types of analyses conducted. We then move in NVivo and import and organise a range of qualitative data. We learn the key features that support a literature review so sources can be annotated and cross-referenced to highlight a line of arguments and connections across sources.

Our attention then turns to the transcribing possibilities of NVivo, starting with transcribing media recordings in-full or working only with sound and video sequences. Working with still images is further introduced. We see that one can work directly on pictures or generate a log to associate comments with specific picture regions. We move on and create externals that link a NVivo project to outside information, as well as creating memos where the analytic process is recorded. Module 1 concludes with lexical queries which search for frequency, occurrence, and context of keywords in textual data.  We analyse the outputs using word clouds, dendograms, and wordtrees.

Module 2: Data Coding          

Module 2 introduces the different techniques to autocode and code data inductively in NVivo. We start by autocoding questions from structured interviews, so the responses of each question are gathered in one node. Such data sorting - known as broad-brush coding - is very useful when one wants to examine everything that was said about a question or a theme across a dataset without having to open each and every source.

We move on with inductive coding and learn the different tools to code data manually. Key notions underlying the coding process such as coding unit, semantic exclusiveness, semantic exhaustiveness, and coding cooccurrence are discussed and exemplified. The use of relationship nodes is introduced to formalise relationships between codes when working towards hypothesis generation or falsification. Module 2 concludes with visualisations that support the coding process from inception to the end.  

Module 3: Data Analysis

Module 3 covers the range of functionalities to prepare and conduct qualitative analysis. Since a large number of social research gather qualitative data, as well as variables, so comparison can be made across cases and sub-sets of cases, we first look at the procedures to create cases from interview data, import variables from Excel, and merge these to the cases. We extend our use of cases to policy documents where comparisons are made on document data, and not cases of individuals. In both instances, we use the functionality of source and node classifications to define type of sources and cases in the dataset.

With the cases created, we turn to the NVivo search tools that efficiently retrieve cases that match a specific search string. This allows us to create sets of cases and documents for comparative analysis.  

We then move on with coding-based queries which retrieve data based on codes overlap, proximity, sequence, or exclusion patterns. We first run coding queries that search for data coded at some nodes but only when mentioned by cases of a given profile. For cross-case analysis, we run matrix queries which cross-tabulate cases with codes, and we interpret the results using different numerical readings: coding density, number of cases, relative percentage, etc. Our interpretation is recorded in memos and is linked back to theory. Module 3 concludes with running group query to find out association between coded items across a dataset.  

Module 4: Data Visualisation

Module 4 proposes different graphic displays to effectively communicate one’s research findings. We first discuss the rationales for choosing certain displays against others. We learn to generate maps, charts, diagrams, and dendograms. Moving on to building a solid audit trail to back up results and substantiate one’s claims, we learn how to export qualitative findings out of NVivo, so these can be used in Word, Excel, and PowerPoint. The usefulness of generating nodes summary reports, which provide detailed synthesis of the scope of a node in a project, is also covered. When working with colleagues who don’t use NVivo, the possibility to export project data in mini websites using HTML files is presented.

Module 4 concludes with the ABC of coordinating team work, with a particular emphasis on the golden rules for successful data management, splitting and merging project files in a master project, and the measurement of intercoder reliability.

No prerequisite knowledge of NVivo required.  Knowledge of qualitative research is necessary.