How to Conduct a True Experiment

Co-authored by Bess Ruff, MA

Last Updated: May 6, 2021 References

Experiments are vital to the advancement of science. One important type of experiment is known as the true experiment. A true experiment is one in which the experimenter has worked to control all of the variables except the one that is being studied. In order to accomplish this, true experiments make use of random test groups.^{[1]
X
Trustworthy Source
PubMed Central
Journal archive from the U.S. National Institutes of Health
Go to source} True experiments are useful for exploring cause and effect relationships such as: is a particular treatment effective for a medical condition? Or, does exposure to a particular substance cause a certain disease? However, because they take place in controlled circumstances, they don’t always fully reflect what will happen in the real world.

Part 1

Part 1 of 3:

Designing the Experiment

1
Formulate the question you would like to answer. Frame your question using the language of cause and effect. Does better nutrition cause higher test scores? Can aspirin cause a reduction in the symptoms of depression?
2
Identify the dependent variable. This is what you hope to change through the experiment. If you are looking for cause and effect, in other words, this is the effect.^{[2]
X
Research source}
- For example, if you want to know if listening to punk music makes you sleep less, the dependent variable will be the numbers of hours slept.
- A dependent variable must be measurable.
Advertisement
3
Identify the independent variable. The independent variable is the factor that you think will cause a change in the dependent variable. It can be thought of as an intervention or a treatment.^{[3]
X
Research source}
- In your cause-and-effect question, it is the term that comes before "cause": does better nutrition cause higher test scores? Better nutrition is the independent variable, and higher test scores is the dependent variable.
- In the example about punk music, listening to punk music is the independent variable.
4
Identify the relevant population. Do you want to study the members of a particular group, such as a college or a city? Are you interested in all diabetic adults, or post-menopausal women, or children who have moved at least twice?
5
Select subjects for the study from your population. If your population is small (for instance, one high school), you might be able to study the whole population. Otherwise, you will need to select a random sample.^{[4]
X
Research source}
- Random selection ensures that your subjects have a diverse set of characteristics that reflects the population in general. This helps you to avoid introducing unintended variables. If education level is significant to your study, for example, and your population includes people with very little education as well as people with Ph.D.s, you don’t want a subject group composed only of college freshmen.
- There are several methods of randomly selecting subjects. For a relatively small population, you could assign each member a number and then use a random number generator to select members. For a larger population, you could take a systematic sample (for example, the second name on each page of a directory) and then use the random number method just described with that smaller subset.^{[5]
  X
  Research source}
- Additionally, large populations can be randomly sampled through stratified sampling methods, which divide the population into homogeneous "strata" and then select individuals from each group to generate a random sample population.^{[6]
  X
  Research source}
- Select a group large enough to produce statistically useful data. The ideal size will vary greatly depending on factors such as the size of the underlying population and the expected size of the effect.^{[7]
  X
  Trustworthy Source
  PubMed Central
  Journal archive from the U.S. National Institutes of Health
  Go to source} You may use a sample size calculator to aid in determining a target size.

Part 2

Part 2 of 3:

Running the Experiment

1
Randomly assign subjects into two groups. One group is the experimental group, while the other is the control group. You must guarantee that any given subject has an equal chance of being in either group.
- Use a random number generator to assign a number to each subject. Then place them in the two groups by number. For instance, assign the lower half of the random numbers to the control group.
- The control group will not be given the treatment or intervention. This will allow you to measure the effect of the intervention.
2
Ensure that subjects do not know which group they are in. If this condition is met, you are conducting what is often called a “single-blind” study.^{[8]
X
Research source} This helps to keep your two groups identical in all respects except the actual intervention or treatment, and is part of controlling for extraneous factors. All members of your study, regardless of group, should believe equally that they are receiving the real intervention or treatment.
3
Ensure that experimenters also do not know which subjects are in which group. If neither the subjects nor the experimenters know, during the experiment, which group is which, you are conducting a double-blind study.^{[9]
X
Research source} This is another way to remove possible extra variables that could affect your study. If experimenters don't know which group is the control group, they won't be able to inform them unconsciously by, for example, administering the inert treatment less carefully.
- Have different people in charge of assigning subjects to a group, administering treatment, and evaluating subjects after treatment.
4
Conduct a “pretest.” In other words, measure the dependent variable before the experiment begins. This can be described as a “baseline” measurement.
- A pretest is not a required feature of the true experiment. However, it increases the ability of your experiment to demonstrate cause and effect.^{[10]
  X
  Research source} In order to say that A causes B, you want to show that A happened before B, which can only be done through the use of a pretest.
- For example, if you are conducting an experiment on how listening to punk music affects sleep, you’d want to gather data on how long each participant typically sleeps at night when they haven’t listened to punk music.
5
Administer the treatment to the experimental group. Ensure that the only difference between the experience of the experimental group and the control group is the treatment itself.
- In a clinical trial, this often means that a placebo is administered to the control group. A placebo resembles the real treatment as closely as possible, but is in fact designed to have no effect. For example, in a study on the effect of a medicine, both groups would come to the same room and receive an identical-looking pill. The only difference would be that one pill would contain the medicine, while the other would be an inert “sugar pill.”
- In other kinds of experiments, keeping the two experiences equivalent will take other forms. Take the example of the effect of playing the trumpet on academic performance. You might want to offer the control group another kind of lesson or opportunity for socialization, to be sure that it’s really the trumpet-playing in specific and not getting a music lesson in general that is causing the effect.^{[11]
  X
  Research source}
6
Administer a post-test. After the course of treatment or intervention is complete, measure the dependent variable. If you conducted a pre-test, the post-test should mirror the pre-test as much as possible, so that the results are directly comparable.

Part 3

Part 3 of 3:

Analyzing Your Results

1
Calculate descriptive statistics. These are statistics that allow you to communicate your data effectively.^{[12]
X
Research source} They provide information about the properties of the data you've produced and allow your readers to understand important things about it from a single glance.When you say, for example, that on average people who received the medicine got better 1.7 days sooner, you are presenting descriptive statistics.
- What is the central tendency of the data? Central tendency is measured using mean (average), median, or mode. For example, in a study on the effects of caffeine on sleep, you will want to calculate the mean number of hours slept by members of the control and experimental groups.
- What is the distribution of the data? Again, there are many different ways to measure how the data are distributed, including range, variance, and standard deviation.
2
Compare the post-test results produced by the experimental and control groups. In addition, if available, compare pre-test and post-test results. To do this, you will need to conduct a statistical analysis of your data. While this is a broad subject, you can make a good start by calculating basic descriptive statistics and by running a t-test to assess if differences observed are significant.^{[13]
X
Research source}
3
Test your hypothesis. Tests of significance will allow you to estimate how likely it is that your results were produced by chance rather than a genuine experimental effect.^{[14]
X
Research source} It determines whether there is a statistically significant difference between the results for the control and experimental groups.
- A t-test is a common test of significance. A t-test compares the difference between the means of two sets of data in relation to the variation within the data.^{[15]
  X
  Research source} You can calculate a t-test by hand or by using statistical software such as Microsoft Excel.
4
Evaluate your experiment. What limits were there on your ability to control possible extraneous factors? To what extent did your subject group reflect the larger population you hoped to study? What alternative hypotheses could be sustained on the basis of your data? When presenting your results, consider limitations honestly and use them to suggest further lines of research.

Community Q&A

Question

Reasons why we conduct laboratory experiments?

Community Answer

Curiosity. Scientific experiments are designed to rule out hypotheses and theories that are wrong. This is done by using the "scientific method", which relies on both inductive reasoning and deductive reasoning; deductive reasoning starts with a general idea and works down to be more specific. Theory > hypothesis > observation > confirmation. This method is used to test theories and hypotheses and is used in very rigorous laboratory experiments.

Warnings

Be sure to take ethics into consideration when conducting this type of study. Never administer anything that may be harmful to a subject. Always stop the study if adverse effects occur. Never withhold treatments knowing that they will improve a subject's health. Follow the guidelines of your school, university, lab, or company in handling human or animal subjects.

⧼thumbs_response⧽
Helpful 0 Not Helpful 0
Be aware of how research design affects results. Bias in how you select subjects or how you control the environment of the experiment can introduce hidden effects on your results.

⧼thumbs_response⧽
Helpful 0 Not Helpful 0