Beginner’s guide on Quantitative Methods in Social Work Research
Updated on Jan 28, 2022
Welcome to graduate school! And welcome to the world of social work research. Social work is fundamentally involved in multi-level systems, and social work research is also interdisciplinary for this reason. This is a good thing. However, you may get lost on which statistical methodologies to learn primarily for that reason, and at least I did! I had questions like these: In what order and how can I learn quantitative methods suitable for my research questions? What types of research questions could be answered by which methodologies? What are the differences between the terminology and methodology used in different disciplines? What software should I learn to use? I hope someone who is lost as I was will find some help in this article.
Choosing your methods: “What do I need to study out of the whole new world of quantitative methods?”
Learning Basic statistics (descriptive statistics, bivariate analysis, regression analysis)
The first step is to learn basic statistics (descriptive statistics, bivariate analysis, regression analysis), starting from the type of variables. It is essential to tell continuous, categorical, and dichotomous variables apart and convert from each type of variable to another type of variable. For example, age could be a continuous variable. It can be converted into a categorical variable for age groups (1=less than 18, 2= 18~30, 3= 31~50, 4=51 and over) or dichotomous variable (0=less than 18, 1= 18 and over), depending on the research questions.
The type of variables decides the type of analyses that you need to choose. So then, you need to learn how to learn bivariate analyses (t-test, chi-square test, correlation analysis, ANOVA, …). The next step is to learn regression analysis. UCLA IDRE guides on “Choosing the Correct Statistical Test” which explains what fundamental kind of primary analysis is appropriate for the independent variables/dependent variables in the research questions. So, you first need to understand the type of the variables (interval, ordinal, continuous, binary, categorical..).
After learning these basic statistics, you can choose to develop your methodological toolkits by choosing your methods depending on your research interests. For example, understanding logistic regression is the most basic step in applying machine learning for the classification task.
Resources for introduction to statistics
MOOC
- Introduction to Statistics (Coursera): https://www.coursera.org/learn/stanford-statistics#syllabus
- UT Austin Statistical Methods in Psychology: https://extension.utexas.edu/statistical-methods-psychology-14320
Book
- Statistics for The Behavioral Sciences by Frederick J Gravetter, Larry B. Wallnau · 2016
- Statistics in Social Work: An Introduction to Practical Applications | Columbia University Press
- Naked Statistics: Stripping the Dread from the Data | Charles Wheelan | W. W. Norton & Company
- Social Work Research Methods: From Conceptualization to Dissemination: Drake, Brett, Jonson-Reid, Melissa
Choosing your methods: History of the development of methods in social sciences
- Psychologists and economists have independently developed statistical methods for social science (Goldberger, 1971). However, they sometimes overlap each other (e.g., some psychologists apply causal inference methods to psychometrics, and some economists apply psychological predictors/outcomes to econometrics, especially behavioral economists).
- Sometimes, statistical terms widely accepted in the field vary by field, but statistical logic is the same. For example, epidemiologists call the interaction effect an effect modification. Sociologists would probably refer to a survival analysis by event history analysis.
- Researchers in public health and epidemiology are more likely to use econometrics or biostatistics (e.g., survival analysis) since the “causality” factor is crucial to their findings. If you are interested in the effectiveness of interventions, causal inference and experimental research design to prove causation before/after intervention are essential.
- Ref: Jeffries, N., Zaslavsky, A. M., Diez Roux, A. V., Creswell, J. W., Palmer, R. C., Gregorich, S. E., … & Breen, N. (2019). Methodological approaches to understanding causes of health disparities. American journal of public health, 109(S1), S28-S33.
- Columbia Epidemiology – List of Public Health Methods
- Social work or education scholars are more likely to use psychometrics since they are interested in subjective concepts considering populations of interest. However, there are exceptions: social welfare policy scholars, educational policy scholars, and interventionists tend to use econometrics more than psychometrics in their research. In recent years, both lines of statistical methods have been developed by integrating computational approaches (e.g., introducing SEM for machine learning).
Psychometrics
- If you are interested in subjective variables (e.g., mental health, stigma, well-being…)
- Utilize psychometric measurement by surveying or collecting a primary dataset
- Structural equation modeling (SEM) is a combination of psychometric measurement (measurement model part) and a series of regression (structural model part)
Econometrics
- If you are interested in (relatively) objective variables (e.g., policy outcome, demographic outcome…)
- Primarily utilize secondary data and observational data.
- Causal inference became the norm in econometrics! Please refer to the methods such as instrumental variable, the difference in difference (DID), and propensity score matching (PSM).
- Some great resources if you are interested in learning econometrics and causal inference methods:
What’s next? Other advanced methods!
There is a line with advanced methods that social work scholars could use; Longitudinal analysis (a.k.a. panel analysis in econometrics), Multilevel modeling, Social Network Analysis, Geospatial analysis, Computational analysis (e.g., Natural Language processing, machine learning …). The methods could be combined and referred to as such, for example, multilevel SEM, longitudinal SEM, and causal inference with spatial data …
Longitudinal Modeling
You may be interested in the longitudinal change that occurs in an individual. For example, you may be wondering how Adverse Childhood Experiences (ACEs) affect life in their later life. To do this, you need to study longitudinal analysis. The methodology you will use may vary depending on whether you are doing psychometric-based or econometric-based research. For example, psychometrics may use longitudinal structural equations (e.g., latent growth models), but econometrics may use the methods for panel analysis (e.g., Generalized Estimating Equations-accounting for non-normality).
Multilevel Modeling
You may be interested in multi-level systems; in social work, Bronfenbrenner’s ecological systems theory is the primary approach that understanding the issues of populations. In many social science research questions, variables are clustered in the upper level of the system. For example, you may wonder about the effect of a particular neighborhood, school, or country-level variable on an individual (e.g., neighborhood effect). It could be answered better using multilevel modeling (also known as hierarchical linear modeling).
Geospatial Analysis & Geographic Information System
Some social work scholars interested in community-level also utilize Geographic Information systems to visualize neighborhood-level characteristics (e.g., community asset mapping).
Reference
- Hillier, A. (2007). Why social work needs mapping. Journal of social work education, 43(2), 205-222.
- Weng, S. S. (2016). Asset mapping for an Asian American community: Informal and formal resources for community building. Psychosocial Intervention, 25(1), 55-62.
- Dunlop, J. M., Chechak, D., Hamby, W., & Holosko, M. J. (2022). Social work and technology: using geographic information systems to leverage community development responses to hate crimes. Journal of Technology in Human Services, 40(3), 201-229.
Social Network Analysis
Social workers care about social relationships and social connections. The Social Network Analysis (SNA) technique is particularly useful if you want to understand how different actors and systems are connected. Duke Network Analysis Center is famous for social networks and health workshops, and fortunately, the workshop materials are now open to all interested in learning SNA.
Reference
- Rice, E., & Yoshioka-Maxwell, A. (2015). Social network analysis as a toolkit for the science of social work. Journal of the Society for Social Work and Research, 6(3), 369-383.
- Hurtado-de-Mendoza, A., Serrano, A., Gonzales, F. A., Fernandez, N. C., Cabling, M. L., & Kaltman, S. (2016). Trauma-exposed Latina immigrants’ networks: A social network analysis approach. Journal of Latina/o psychology, 4(4), 232.
Data Science / Computational Social Science
If you are interested in using big data in particular (e.g., electronic health records or administrative data in child welfare or justice system) or utilizing text as a quantitative form of data (e.g., social media posts), computational methods or data science can be the methods you should learn. Natural Language processing is the technique that transforms text data into a computable format.
Others
- If you are interested in applying the intersectionality approach in the quantitative method, please refer to this article: Guan, A., Thomas, M., Vittinghoff, E., Bowleg, L., Mangurian, C., & Wesson, P. (2021). An investigation of quantitative methods for assessing intersectionality in health research: A systematic review. SSM – Population Health, 16, 100977.
Statistical software
Since replicability and collaboration are getting increasingly important in social scientists’ research (lots of journals encourage the authors to provide the data and code in submission), I recommend you start to practice coding instead of using SPSS or AMOS as soon as possible. In social sciences, Stata is most widely used. Stata free webinars can be found here: https://www.stata.com/training/webinar/
- Coding experience is required in the following order: SPSS (requires no coding) > Stata (requires coding for replicability, but it can be done without coding) > R (requires a lot of practice in getting accustomed to the package)
- For structural equation modeling: AMOS (requires no coding) > Stata (requires coding for replicability, but it can be done without coding, but AMOS and Mplus provide more options for SEM analysis) > Mplus (requires coding but not extremely difficult) > R (requires lots of coding)
Recently, R has been gaining popularity since it is free and can be expanded by user-written packages! Once you get comfortable with writing the code (e.g., Stata), the speed of learning how to code with another software (e.g., R) can be accelerated. It is your choice to start coding from R or Python instead of Stata. Still, I think you can start from any syntax-based program (which I mean by requires coding) since familiarizing yourself with syntax writing is the asset that makes you learn another software with less anxiety and frustration.
Update: ChatGPT even provides the functionality that converts the code from one language (e.g., STATA) to another one (e.g., R). If you are comfortable with any language, it will get easier for you to do it in another language.
Workshops/Resources
Outside of school, advanced statistical methods can be learned from these websites (some of them are too expensive to pay out of your pocket):
- ICPSR Summer Program
- Episummer@columbia
- Centerstat workshops
- The Analysis Factor Workshops
- UCLA IDRE webinars and notes
- UCLA IDRE upcoming webinar (Live & Free)
- Workshops @ Institute for Social Science Research | UMass Amherst
- Mplus short courses (for SEM)
- Quantfish Workshops
- Johns Hopkins Graduate Summer Institute of Epidemiology and Biostatistics
- Dr. Laura Bronner, Quantitative editing guide
- Dr. Lesa Hoffman, Course Materials (Free)
- You can find the syllabus with videos on a variety of methods, including Longitudinal Multilevel Models, Clustered Multilevel Models, Factor Analysis and Structural Equation Models, Generalized Linear Models, and Latent Trait Measurement and Structural Equation Models.
If you are interested in cross-cultural research and structural equation modeling, I have some recommendations on books.
- Tran, T. V., & Chan, K. T. (2021). Applied Cross-cultural Data Analysis for Social Work. Oxford University Press.
- Tran, T. V., Nguyen, T. H., & Chan, K. T. (2016). Developing cross-cultural measurement in social work research and evaluation. Oxford University Press.
- Kline, R. B. (2015). Principles and practice of structural equation modeling. Guilford publications.
1 Response
[…] Beginner’s guide on Quantitative Methods in Social Work Research […]