Skip to Main Content
It looks like you're using Internet Explorer 11 or older. This website works best with modern browsers such as the latest versions of Chrome, Firefox, Safari, and Edge. If you continue with this browser, you may see unexpected results.

Social Work Research Guide

An introduction to finding Social Work Resources

Levels of Evidence

A pyramid graphic showing the different levels of evidence in research

Research Methods

A mind-map of different research methods

Study Design

Study design: What are the basic categories of research design? The classifications that can be used include Observational versus Experimental, Prospective versus Retrospective, and Longitudinal versus Cross-sectional. The distinction between observational and experimental relates to the purpose of the study, and is the most important distinction. The other two distinctions relate to the way in which the data are collected. Most of the combinations of these terms are possible, although not all are. These are the basic types of research design. Some of these categories are described in more detail in subsequent sections.

Types of correlational research

There are many kinds of correlational research in which the interrelationships of pairs of variables are explored. Causal-comparative studies (either prospective or retrospective) compare two or more groups of subjects on one or more variables in order to determine whether or not there is a ‘case’ for causality.

Study Design


The design of the study refers to the means by which the research question will be addressed, specifically in relation to the data that will be collected, the comparisons that will be made, the experimental conditions (if any) that will be manipulated, and so on. In many ways, the design of a study is more important than the analysis of the data resulting from that study; if data are poorly analyzed it will always be possible to reanalyze them, but if a study is poorly designed in the first place then it may never be possible to meaningfully interpret the data which result from it. The design of a study will also have an impact on how data are subsequently analyzed.


Research can be defined as belonging to one of two primary categories: observational studies and experimental studies. As suggested by the name, in observational studies there is no direct manipulation of variables within the study, and data are simply collected on groups of participants. In an observational study, while the researcher collects information on the attributes or outcomes of interest, no effort is made to influence these. An example might be the prevalence of a particular health behavior (e.g. cigarette smoking) in different groups defined by socioeconomic status. In an experimental study, on the other hand, the experimenter directly influences events, in order to draw conclusions regarding the impact of that manipulation on the resulting observations. An example might be the impact of an intervention (e.g. an increase in self-efficacy beliefs) compared to a control condition on a particular health behavior (e.g. giving up smoking).

A further sub-division which may apply to observational studies is between prospective studies, where information is collected about events subsequent to recruitment into the study, and retrospective studies, where information is collected about events in the past. Data regarding past events may be obtained from more objective sources (e.g. school records or patient notes) or, arguably, from more subjective sources (e.g. by self-report by the participant). Note that, in all cases, experimental studies are prospective (even if the period of prospective study may be very short!), whereas observational studies may be retrospective or prospective. It should also be noted that whereas different treatments can be assessed retrospectively, this would not count as an experimental study, since the delivery of different treatments would not be an element of a pre-specified study. In this case, the study would be a retrospective observational study.

Another sub-division is that studies may be classified as longitudinal studies or cross-sectional studies. In longitudinal studies, changes over time are investigated. Again, note that, in all cases, experimental studies are longitudinal (although once again, the period of longitudinal study may be very short), whereas observational studies may be longitudinal or cross-sectional. In a cross-sectional study participants are observed only once, offering a 'snapshot' of the characteristics of interest at that particular moment. Most surveys, for example, are cross-sectional, although a variant is the pseudo-longitudinal design, where data on participants are collected at only one point in time, but this is done, for example, on, participants of different ages in order to indirectly construct a distribution of changes in a population with age. This approach is usually less resource-intensive than a proper longitudinal design, although somewhat more prone to giving erroneous results.

RESEARCH METHODS AND MEASUREMENT AND: STUDY DESIGN. (2008). In I. P. Albery, & M. Munafo, Key concepts in health psychology. London, UK: Sage UK. Retrieved from

One of the three major research paradigms, which includes qualitative research, mixed methods research, and quantitative research. Quantitative research relies primarily on the collection of quantitative data and has its own, unique set of assumptions and normative practices. Quantitative research usually assumes that human behavior exhibits some lawfulness and predictability that can be documented through empirical research. Goals include to describe, to predict, and to explain human phenomena. Quantitative researchers often try to study behavior under controlled conditions via experiments, in order to isolate the causal effects of independent variables. Popular methods of quantitative research are experimental research, survey research, and structured observational research. Quantitative data are collected based on precise measurement of variables using structured, standardized, and validated data collection instruments and procedures. Data are analyzed using descriptive and inferential statistics. The desired product is research findings that generalize broadly. Quantitative researchers attempt to minimize their biases during the research process by relying on standardized testing and measurement, continual testing of measurement procedures for reliability and validity, random selection of research participants, random assignment of research participants to comparison groups, measures of interrater reliability when multiple observers are used, and inferential statistics for estimating the values of population parameters and for testing statistical hypotheses. 

Quantitative research. (2009). In L. E. Sullivan (Ed.), The SAGE glossary of the social and behavioral sciences. Thousand Oaks, CA: Sage Publications. Retrieved from

Qualitative research goes under several sobriquets, such as the humanistic model of social research, unobtrusive methods and ethnographic approaches. ‘Qualitative research’ is the preferred term for most people. It is a term used to describe an approach to research than stresses ‘quality’ not ‘quantity’, that is, social meanings rather than the collection of numerate statistical data. For example, qualitative research might explore how an individual who voted for the Green Party in the United Kingdom sees themselves as a member of a minority party dealing with environmental issues rather than, say, exploring overall voting trends over time for the Green Party and other minority parties within Britain's two-party system. It is normally contrasted with quantitative research, an approach to research that stresses the reverse dimension. For example, if one wanted to explore what being vegetarian means to someone, the focus would be on the social meanings around organic food use and animal welfare and so on, resulting in the use of qualitative research. If one wanted to know how many people in an area were vegetarian, and perhaps voted for the Green Party as well, one would use a questionnaire and social survey to generate numerate data. Both kinds of data are valuable for their respective purposes.

Willig, C. (2016). Qualitative research. In L. H. Miller (Ed.), The Sage encyclopedia of theory in psychology. Thousand Oaks, CA: Sage Publications. Retrieved from

A literature review is typically the section of a research paper, dissertation or thesis that considers the writing that other authors have already produced on the topic. A literature review may also occasionally be referred to as a literature search or a critical literature review. This latter term or label highlights that the central aim of a review is to critique, analyse, compare and contrast various writings on a given area.

Although not obligatory, it is often a good idea to commence your literature review with the literature dealing with some of the background or historical context to the issue(s) being considered. This sets the scene and allows the review to lead on the literature that goes on to address the contemporary situation and issues in relation to that background context.

Literature review. (2011). In P. Stokes, Palgrave Key concepts: Key concepts in business and management research methods. Basingstoke, UK: Macmillan Publishers Ltd. Retrieved from

A systematic review is a comprehensive review of literature which differs from a traditional literature review in that it is conducted in a methodical (or systematic) and unbiased manner, according to a pre-specified protocol, with the aim of synthesising the retrieved information through meta-analysis, often using statistical tests. A systematic review can be considered analogous as primary research where the cases are research publications. The reviewer must specify: how the publications (cases) will be selected; the type of instrument that will be used to obtain data from the publications; the methods to be used for this data collection; and the type of analysis that will be conducted on the data. Therefore, when undertaking a systematic review, the researcher must follow an explicit path. This will be outlined, with reference to an example of a systematic review conducted by Thomson, Petticrew and Morrison (2001).

DEMPSTER, M. (2003). Systematic review. In R. L. Miller, & J. D. Brewer, The A-Z of Social Research. London, UK: Sage UK. Retrieved from

One of the earliest and certainly the most influential definitions of evidence-based practice (EBP) is by Sackett et al. (1996): 'Evidence-based medicine is the conscientious, explicit and judicious use of current best evidence in making decisions about the care of individual patients.' The focus of this definition is clearly on the use of research to inform decision-makers and enhance their decision-making capabilities. The decision-makers appear here to be the clinicians, those responsible for the care of individual patients, and indeed the experience and judgement of clinicians is an integral part of the process. However, an unrepresented group in this definition are patients, users or clients and it is essential that their subjective values, preferences, beliefs and opinions are incorporated into the decision-making process.

DEMPSTER, M. (2003). Systematic review. In R. L. Miller, & J. D. Brewer, The A-Z of Social Research. London, UK: Sage UK. Retrieved from

Meta-analysis is a quantitative approach that permits the synthesis and integration of results from multiple individual studies focused on a specific research question. A meta-analysis is a rigorous alternative to the traditional narrative review of the literature. It involves the application of the research process to a collection of studies in a specific area. The individual studies are considered the sample. The findings from each study are transformed into a common statistic called an effect size. An effect size is a measure of the magnitude of the experimental effect on outcome variables.

Once the results from each study have been converted to a common metric, these findings can be pooled together and synthesized. The most common effect size indicator is r, which is the Pearson product moment correlation. Another effect size indicator is the d index. Cohen's d is the difference between the means of the experimental and control groups divided by the standard deviation. Cohen (1988) has provided guidelines for interpreting the magnitude of both the r and d effect size indicators. For the r index, Cohen has defined small, medium, and large effect sizes as .10, .30, and .50 or more, respectively. For the d indicator, an effect size of .2 is considered small, .5 is medium, and .8 or more is large.

Beck, C. T. (2017). Meta Analysis. In J. Fitzpatrick (Ed.), Encyclopedia of nursing research (4th ed.). New York, NY: Springer Publishing Company. Retrieved from