Research design is the overall plan or approach that a researcher follows in order to study a particular research question. There are many different research designs that can be used, depending on the specific goals and characteristics of the research project.
For example, a researcher might use a descriptive research design to simply observe and describe a particular phenomenon, or a experimental research design to test a hypothesis by manipulating variables and observing the effects. Other common research designs include cross-sectional, longitudinal, and mixed-methods designs. These research designs are used to guide the collection and analysis of data, and to help ensure that the research is rigorous and reliable. The following are common types of research design.
Review and narrative that is based on existing sources.
Analysis that uses existing sources. For example, a review of multiple studies that numerically aggregates and summarizes their findings.
Primary research produces new observations. Also known as original research.
Collecting, analyzing, and interpreting non-numerical data such as interviews with people.
Collecting, analyzing, and interpreting non-numerical data such as sensor readings.
Research that strictly conforms to the scientific method including elements such as a falsifiable hypothesis, empirical evidence and peer review.
Correlational research looks for correlations between variables without manipulating these variables. Correlation doesn’t equal causation such that these studies can produce misleading impressions that one thing causes another when both may be influenced by some third factor.
Using software to automatically find correlated variables in datasets. This can be used to produce fraudulent research whereby a researcher misrepresents their method by pretending to start with a research question when they actually worked backwards from automatically discovered correlations. Data dredging also plays a valid role in exploratory research.
Research that lays the groundwork for other research. For example, a data analysis that is used to formulate a problem statement, hypothesis or experiment design.
Causal-comparative research attempts to use data to establish evidence for a cause and effect relationship. This might use several datasets and detailed controls that aggressively seek to eliminate alternative explanations for an effect. For example, if people who live near busy highways have a higher risk of some health problem a study may control for other factors that may explain this correlation such as income level or lifestyle.
Research where the independent variable isn’t controlled such that it isn’t an experiment. This can be exploratory research, correlational research or causal-comparative research.
Studies based on groups of people who share a common characteristic, known as a cohort.
Choosing the members of cohorts at the start of a study.
Cohorts are selected based on historical data. Runs some risk that the researcher will aggressively define the cohort to fit some pattern found in the data.
A retrospective cohort selected based on outcomes such as comparing the lifestyle of people who get a disease with those who don’t get it. Useful for exploratory research but problematic for establishing cause and effect. For example, if you scan for differences in the lifestyle of people who graduate high school and those who don’t you may find that jelly donut consumption are different between these two cohorts but it is a stretch to suggest this is a cause.
A detailed report of a single example. Useful for exploratory research. For example, a doctor who documents an allergic reaction to a chemical that hasn’t been on the market for long.
Measuring the same variables over an extended period of time. Often an observational cohort study that observes a group of people over some time period. However, experimental research can also be a longitudinal study such as an experiment on a field of crops for half a year.
A study that compares observations at a point in time. For example, comparing the air quality of cities and the rate of a disease in those cities with the most recent data available for each city.
Experimental research is the testing of a hypothesis or multiple hypotheses with experiments. This involves changing an independent variable to observe corresponding changes to a dependent variable. For example, a researcher who produces different types of stainless steel formulations to test which is most resistant to seawater.
An experiment in a lab where many variables can be controlled. For example, testing a fertilizer on plants in a lab where you can control extraneous variables such as light, temperature, humidity and water.
An experiment that occurs in the real world where some variables can’t be controlled. For example, testing a fertilizer on a farm.
Randomized Controlled Trial
A standard for important experiments such as clinical trials for medical treatments that uses random allocation of participants to treatment and control groups to achieve statistical control over factors that may influence results. For example, if body weight may influence the outcome of a trial, people can be randomly distributed into treatment and control groups such that body weight distributions are likely to be similar in each group.
A natural experiment is a real world situation that resembles an experiment. This is useful were experiments would likely be unethical. For example, a factory where workers are currently exposed to a hazardous substance.
Constructive research addresses a real world problem. For example, computer science research that seeks to design algorithms to perform a computation more efficiently.
Research & Development
Constructive research that designs a process, method, procedure, device, machine, product or service. For example, rapid prototyping of possible battery technologies.