Skip to main content

Material for HDR candidates

Choosing your research methodology and methods

Photo of magnifying glass

The most challenging activity in many research projects revolves around deciding which methodologies and methods to utilise. This set of resources will help you

  • decide which theoretical perspectives, methodologies, and methods you should use 
  • decide which techniques you could use to analyse data
  • identify other methods that could improve your research

Once you choose a specific methodology or method, you might need to seek more advanced information about these choices.  For example, you could read suitable books or complete online training courses.  Sometimes, the most effective courses may cost several hundred dollars--such as the courses available from the Australian Consortium for Social and Political Research.  You may decide to utilise your funding allocation to attend one or more of these courses.  


Choosing qualitative approaches, methodologies, and methods

During many research projects, candidates will collect and analyze qualitative data--often words instead of numbers. If you need to collect or analyze qualitative data, you must then decide which approaches, methodologies, and methods to utilize.

  • read this document to chose suitable approaches, methodologies, and methods
  • also skim the potential alternatives in the following table


This table presents more information about specific examples of qualitative--as well as mixed method--methodologies and methods


When this topic is important

Document to read

Theoretical paradigms




Researchers often adopt this approach if they want to collect both qualitative data and quantitative data

Decolonising theory

Researchers often, and perhaps should adopt this perspective if they want to conduct research with Indigenous communities

Critical race theory

Researchers often adopt this perspective when they want to redress discrimination towards specific races or ethnicities

Standpoint theory

Researchers often apply this theory to show that research conducted by dominant collectives, such as affluent men, tend to overlook or distort the experiences of women or marginalized communities. This theory is often applied to justify feminist or Indigenist research

Methodologies: Classical  
Grounded theory

You want to develop a theory about why people engage in certain social behaviours in a particular context--and want to test and refine this theory iteratively.  To learn about a specific variant in detail, read


Narrative approachesYou want to analyse people's stories about their lives or construct stories about people's lives
Interpretative phenomenological analysis and other similar approachesYou want to help interesting people share their lived experience--their perspective or perception of their world
Participatory action researchYou want to collaborate with a community of individuals to solve a problem that is of interest to this community
Case studies, as recommended by Robert YinYou want to explore one or more intriguing cases--such as a fascinating person, event, organization, practice, or intervention--in context, using a variety of methods and techniques
Mixed methods researchYou want to collect both qualitative data, such as interviews, and quantitative data, such as numerical information

Methodologies: Ethnographies



Policy ethnography

You want to explore and characterise the experience of individuals, such as community health managers, who need to implement some government policy. You might, for example, want to uncover the obstacles and challenges that impeded the implementation of some policy--and should be addressed in the future.


You and another researcher want to analyse and report your conversations with each other about a shared issue in your life, such as counselling violent people or the development of sexual orientation.

Methodologies around language



Critical discourse analysis

You want to explore how the words, phrases, and images that authors or speakers utilise may sustain or amplify disparities in power and, therefore, promote injustices

Conversation analysis

You want to understand how people use conversations and language to achieve social goals, such as to convince, persuade, or change other individuals

Program evaluations


Utilisation-focussed evaluations

You want to evaluate some program, initiative, or intervention--but you are willing to work collaboratively with stakeholders to increase the likelihood your evaluation will be heeded

Outcome harvestingYou want to first identify favourable outcomes of various programs--and then clarify which programs, initiatives, or practices explain these outcomes.  This approach is suitable when many programs or practices may be contributing to various outcomes

Realist evaluation

You want to determine which people benefit from some complex, social intervention--comprising many features and activities--and the circumstances in which this intervention is beneficial

Appreciative inquiry

You want to uncover actions a team, organisation, or community could initiate to improve their conditions--but derive the ideas from positive experiences in the past

Cost effectiveness analysisYou want to establish whether the benefits of some program, initiative, or intervention outweigh the costs.  Useful if the benefits cannot be readily monetised 
Social return on investmentYou want to determine whether the social, environmental, and economic benefits of some program or intervention outweigh the costs
Unorthodox methodologies  
Actor network theoryYou want to explore how humans, documents, objects, artefacts, and other materials all affect one another to construct, modify, sustain, and dissolve some association, event, or phenomenon

Horizon scanning

You want to develop plans or practices that accommodate possible problems or opportunities that might unfold in the future

Process evaluation

You want to evaluate how an intervention was implemented--rather than merely whether some intervention was effective. For example, you might want to assess why certain protocols were not observed, the effects of these deviations, and so forth.

Systematic intervention constructionYou want to integrate all the practical implications that have been recommended about a topic--such as how to reduce weight or alleviate depression--to construct a unified program or intervention
Data collection: Variations on traditional approaches  
Dyadic interviewsYou want to interview two participants at a time, to exploit the benefits, but avoid the complications, of focus groups
Episodic interviewsYou want to encourage participants to recall specific events, moments, or experiences that are relevant to your research topic--as a means to explore this topic in more depth
Observational researchYou want to observe some event or circumstance as systematically and ethically as possible
Qualitative social network analysisYou want to explore how and why information, resources, friendships, and other attributes spread across people, organisations, or communities—but want to apply a qualitative approach to explore these relationships in detail

Data collection by participants



Systematic self-observation

You want to teach participants to collect data about themselves in specific circumstances, such as when they tell lies

Participatory visual methodsYou would like participants to use cameras to record their lives--or use photos, videos, drawings, and other artefacts to elicit memories during interviews
Web-based diary studiesYou would like participants to maintain a diary, primarily to record, and then share, their experiences while they unfold

Data collection from the web



Research with Twitter

You would like to use information from Twitter to answer your research questions

Web scraping

You would like to extract content from many websites as efficiently as possible--content that you can then subject to a range of analyses later

Data collection from discussions or stories



Story completion methods

You want to ask people to write stories--as a means to uncover the motivations, perceptions, explanations, and attitudes of individuals towards sensitive topics


You want to learn about Indigenous experiences, perspectives, and ideas from conversations with Indigenous people; this document also outlines some relevant paradigms, such as slow science

The nominal group technique

You want to arrange members from some community to identify and rank possible solutions to some problem they share

Data analysis and presentation 



Thematic analysisYou want to extract several key themes, concepts, or insights to understand some phenomenon better; one of the most common techniques to analyse qualitative data.  This document also introduces template analysis--a variant of thematic analysis in which researchers often develop a set of possible themes before they commence the analysis
Framework analysisYou want to uncover themes and patterns from your data--but you want to utilise or integrate previous taxonomies, framework, or knowledge as well.  An alternative to template analysis, but often used in the health sector

Tensional analysis

You want to uncover, explore, and describe the tensions or conflicts that pervade a particular setting or location, such as the tensions that research candidates or teachers experience in modern life

Situational analyses

You want to develop a theory or account of some phenomenon--similar to grounded theory--but want to construct maps or diagrams to appreciate some of the complications, contradictions, and instabilities


You want to convert the data you derive from interviews--data about some community or setting--into a theatrical performance, like a play

Data analysis with software  

Content analysis and concept maps with Leximancer

You want to uncover common themes from various texts, such as emails or books, using an algorithm

Text analysis using RYou would like to numerically examine and analyse the texts you downloaded from the web or other sources

Comparing different texts

You want to ascertain the differences between two sets of texts, such as social media before COVID-19 and social media after COVID-19

Classifying textsYou want to use machine learning to classify texts

Which cluster of quantitative techniques should I use?

During many research projects, candidates will collect and analyze quantitative data--data that involves numbers or counting. If you need to collect or to analyze quantitive data, you first need to decide how to design your research and which cluster or category of techniques you will utilize to analyse the data. The following information could help you answer these questions.

Materials to readContent of these materials
Defines the main research designs
Offers advice on which cluster of techniques you should use to analyse your data
SPSS video on Youtube 

Teaches basic statistics and SPSS. If your statistical expertise is limited, this video will greatly help you understand the other material on statistics and data analysis--even if you do not plan to use SPSS in the future.  You can also use this workbook to help you learn the basics: 

Introduces R--perhaps the most powerful free tool you can use to analyse and to manage your data

Which specific techniques should I use?

Once you have decided which cluster of techniques is suitable, you then need to choose a more specific technique. This table may help you achieve this goal.


When this topic is important

Document to read

Basic analyses



Overview of between-subject analyses

You want to compare groups on one or more measures

Overview of within-subject or mixed-model analyses

You want to examine whether some measure changes over two or more times or to compare matched groups

Overview of correlation or regression analyses

You want to explore the association between numerical variables

Non-parametric statisticsOften useful if the outcome measures are ranks, ordinal, categorical, or violate assumptions, such as normality

Linear or multiple regression analysis

You want to explore the association between one numerical outcome and multiple variables.  If you want to impress the examiner or reviewer and determine the importance of each variable, read this document too

Comparing groups or classifying cases  

Logistic regression

You want to explore whether two groups or conditions differ on a set of measures or characteristics; you can extend this technique to more than two groups as well. If you want to impress the examiner or reviewer and determine the importance of each measure or characteristics, read this document too

ROC curves

A technique that assesses the accuracy of measures or tests that are designed to predict one of two outcomes, such as whether or not someone should be diagnosed with some disease

Latent class analysis

You want to uncover subsets or clusters of participants--participants who are similar on various categorical variables

Advances in regression 



Moderated regression analysis

You want to explore how some other variable affects or moderates the association between one numerical outcome and other variables. For a file to help you construct, the graphs, click

Ridge, lasso, or elastic net regressionYou want to improve your linear regression analysis--and, in particular, increase the accuracy of predictions.  This technique will impress many examiners or reviewers
Fractional polynomial regressionYou want to explore the association between some numerical outcome and various predictors--but at least one of the associations deviates from a straight line and instead is curved or complex
Poisson, negative binomial, and zero-inflation modelsYou want to explore the association between one outcome and multiple variables--but the outcome measure is a count, such as the number of times each participant initiates some act, such as visits a doctor.  In these instances, linear regression is often not applicable

Generalised linear equations

You want to explore the association between one outcome and multiple variables--but the outcome measures are counts, ranks, ordinal, categorical, or violate assumptions, such as normality. Thus, linear regression is not applicable

Multi-level analysis

Useful if your participants, animals, specimens, or other units are derived from several broad clusters that you are not interested in comparing—or should be used in lieu of within-subject or mixed-model analyses when some of the data are missing

Generalized estimating equations

Useful if your participants, animals, specimens, or other units are derived from several broad clusters that you are not interested in comparing--or your design is longitudinal--but the outcome measures are not necessarily numerical or linearly related to your predictors

Generalized additive models

Useful if the association between some predictor, such as time, and some outcome, such as weather, did not follow a regular pattern but shifts quite erratically

Structural equation modelling  
Exploratory factor analysis

Identifies sets of items or questions that are correlated with each other—commonly used to evaluate or develop scales, measures, or characteristics

Mediation modelsYou want to examine the relationship between a sequence of variables--often to explore why two variables are related
Path analysisYou want to examine more than one cluster of relationships simultaneously.  In contrast, linear regression explores only one  pathway that connects the independent variables to an dependent variable
Structural equation modellingSimilar to linear regression or path analysis, but applicable when the key variables are composites of multiple questions, items, or indicators.  This technique is versatile
Growth curve modellingExplores how variables change over time--as well as the conditions or circumstances that affect these changes over time
Advances in statistics  
Basic time series analysis and Google Trends

Useful if you want to examine how some characteristic, such as the popularity of some topic, changes over time.  This document also introduces Google Trends--a webpage that you can use to explore the popularity of various topics over time.  This document refers to the following file


Vector autoregressionUseful if you want to examine how two variables affect each other over time--such as unemployment and consumer confidence

Survival analyses, such as Cox regression

Useful if you want to explore which predictors or circumstances affect when some event is likely to unfold--such as death, completion of a task, or withdrawal from some setting

Social network analysisUseful if you want to explore how friendships, information, resources, associations, and other attributes spread across networks of people, teams, organisations, or communities.
Item response theoryA technique that assesses the suitability of assessments that are designed to measure ability or capability--especially if the level of difficulty varies across the questions

Bayesian statistics: An introduction

Often utilized when you want to integrate past expectations or information with the data you have collected

Machine learning  
K nearest neighbour and machine learningIntroduces the notion of machine learning as well as outlines one technique, called k nearest neighbour--a technique often used to classify and predict outcomes--such as which book someone will like based on their previous purchases
Support vector machinesA variant of machine learning is often used to classify people, animals, or objects into one of several categories; useful when the number of attributes and cases is moderate
Neural networksA variant of machine learning that is useful if you want to predict some outcome from a series of predictors
Decision treesAnother variant of machine learning is also useful if you want to predict some outcome from a series of predictors; unlike neural networks, the outcomes are easier to interpret
Random forestsA more accurate and sophisticated variant of machine learning than decision trees--also useful if you want to predict some outcome from a series of predictors; read the document on decision trees first
AdaBoostLike random forests, useful if you want to predict some categorical outcome from a series of predictors; read the document on decision trees first.  Knowledge about AdaBoost will also help you learn about other machine learning methods 
Gradient boostingLike AdaBoost, but useful if you want to predict either categorical outcomes or numerical outcomes from a series of predictors; read the document on decision trees first.  Knowledge about gradient boosting will also help you learn about other machine learning methods 
Naive BayesA variant of machine learning that is useful if you want to predict some categorical outcome from a series of predictors--and is effective if the predictors are uncorrelated with each other.  Often used to categorise texts

Fundamental issues in statistics

When analyzing data and conducting statistical tests, common questions arise, such as how many participants should you recruit.  To answer these questions, you should first learn about power and effect sizes.

Document to readContent of these documents
Demonstrates how to calculate effect sizes--a statistic that many journals now recommend
Offers some advice on how to determine the appropriate sample size--the number of participants, animals, organizations, fields, specimens, and so forth you need to complete your studies successfully. Read this document only after learning about effect sizes in the previous document
Reveals how to increase the likelihood of significant results


Specialist techniques

Some specialized methods and approaches can be utilised in a variety of circumstances. 

Documents to considerContent of these documents
Introduces multiple criteria decision making. This technique can be used if you want to choose one of several options that differ on multiple criteria--such as choose a suitable teaching method to apply or house to purchase. This technique can be used to decide which methods to utilise or as a method itself.
Introduces sensitivity analysis.  This approach is useful whenever you need to apply some formula or equation to predict some outcome—and then want to demonstrate the accuracy or certainty of this outcome.  
Introduces fuzzy cognitive maps. This technique can be used if you want to predict how some decision, such as a policy in which candidates must work at home, affects various outcomes. But, this technique is especially valuable when all the variables or considerations are likely to affect each other
Introduces Q methodologies. This technique can be used to analyse the opinions of people on some topic.
Introduces the Delphi method--an approach in which you seek the opinions of experts online several times in a systematic fashion
Introduces lead user research--an approach in which you seek the insights of people who have modified and improved a specific product, service, or practice, ultimately to unearth possible innovations
Accommodating complexities in your design 
Introduces propensity score matching--used to improve the accuracy of quasi-experimental designs in which participants are not randomly assigned to conditions
Introduces step-wedge designs--in which you gradually introduce some intervention to more groups or people over time.
Introduces an approach to utilize when comparing intervention to a treatment when the sample size is small. When you apply this approach, you compare each score in the intervention condition with the mean of the control condition.
Techniques related to COVID 

Presents a range of measures and tools that could be useful if you want to conduct social science research that is related, even if obliquely, to COVID-19