Experimental Research Design

An experimental research design serves to identify the causal relationships between independent variables. We use it to measure the effects of one variable on the other or on dependent variables.

You create a row of procedures to check hypotheses and, therefore, need to understand the whole system under research well.

Identify the Variables

Start with a unique research question. Let’s regard the following examples:

Research question 1 (related to health sciences) is developed to investigate the use of phones and sleep quality. The question can the following - How the amount of time a person uses a phone before going to bed influences the number of hours they can sleep after that?

Research question 2 (related to ecology) means to investigate the relationships between the temperature and soil respiration. It can sound like that - How much the rise in temperature above the surface of soil can change the amount of carbon dioxide, which is emitted from the soil.

Now, you can make up an experimental hypothesis by defining the major variables. In this way, you can make predictions about their relationships. Begin with defining your dependent and independent variables. For example, for the relationships between the use of phones and sleep quality, an independent variable can include the number of minutes a person uses the phone before sleep, and a dependent variable is the number of sleep hours per night.

In case of the relationships between the temperature and soil respiration, an independent variable sounds like the temperature above the surface of the soil, and a dependent variable is the amount of CO2 which is respired from the soil with the rise in temperature.

After you have defined these variables, consider all possible extraneous variables as well and think how you can control them during the experiment. For instance, when you research the relationship between the use of phones and sleep, a supplementary variable can be the individual difference in sleep patterns in individuals. Statistical control is recommended here. You can measure the difference in sleep hours between those who sleep with the use of phones and those who do not use them. You should not measure just an average amount of sleep within the group because the result will not be valid.

When you deal with the relationship between the temperature and respiration of soils, a possible extraneous variable is soil moisture, which can influence the respiration or the sufficient decrease in moisture when the temperature increases. To control these variables experimentally, you need to observe the soil moisture carefully and add water every time you see that it has decreased to keep it consistent during the whole time of observation.

Another way is to make a diagram with the variables and use arrows to demonstrate the predicted relations between the variables. You can also indicate the possible directions of such relations.

Make Up a Hypothesis

After you have got the complete idea of the system under research, you can make up a unique hypothesis that can be tested and related to the research question.

As you know, the hypotheses can be null (H0) and alternative (H1) ones. For the use of phones and sleep, the null hypothesis can be the following - the number of hours an average person sleeps do not relate to their phone use. The alternative hypothesis here is that when a person increases the number of hours for using a phone, the sleep amount decreases. When you make a null hypothesis for the temperature and soil respiration, it can imply that the temperature of the air has nothing related to soil respiration. The alternative hypothesis is that the increase in temperature leads to the rise of soil respiration.

Then, you can start designing a controlled experiment. It will allow you to:

• manipulate the independent variables regularly;
• make accurate measurements of the dependent variables;
• take control over all possible confounding variables.

If you cannot apply these recommendations to the system you are studying, look for other types of research that will be more applicable to it.

Treat the Independent Variables According to the Experimental Design

The correct treatment of independent variables can influence the overall validity of experiments. In its turn, such validity will affect the further use of the results.

The first question to answer here is how much to extend the independent variable. For example, in the experiment related to soil respiration and temperature you can opt for increasing the air temperature according to the following patterns:

• slightly above the common temperatures of your region;
• higher than average to imitate the possible future warming;
• much higher above the highest possible range which is unlikely to happen in the natural environment.

The second question relates to the choice of the most accurate variations of your independent variable. Rarely, the experimental system makes this choice itself. However, the most common situation is when you need to opt for the most appropriate variation yourself to be sure that the results will be the most valid ones.

For instance, in your experiment related to the use of phones and sleep, you can treat the phone in the following ways:

• a continuous variable counting the minutes/hours of the phone use per night within a certain period of time;
• a categorical variable that can be either binary (answering Yes/No questions) or leveled implying no use of phones, low use, or high levels of using.

Relate Subjects to Their Groups under Research

If you want to get reliable results from your experiments, you need to correlate the experimental activities with the subjects you test.

To start doing this, define the size of your study or number of the participants. The individuals can also include other living and non-living objects included into the study. The main idea here is that the statistical power of your experiment will be greater if you include more subjects in it. A larger number of individuals means that you will be more confident about the results.

To proceed, you should assign the individuals to the specific groups to ensure the proper level of treatment. In case of situation in this article, it means forbidden use of phones, low levels, or high levels of using. Create a control group too. No activities should be done within it during the experiment. This group will inform you what would happen to any group under research if you did not do anything about it.

To ensure the best outcomes from assigning individuals to their groups, you will have to develop the following things:

• a fully randomized design to contrast the randomized block design;
• a within-subject design that can be contrast to a between-subject design.

How to Maintain Randomization

You can randomize your experiment completely or just make an aka-strata randomization (by blocks):

When the design of your experiment is completely randomized, you assign each subject to a specific treatment group randomly.

When you use a block design, all the subjects are classified according to one common characteristic and then randomly assigned to their specific groups.

Let’s consider these procedures in our examples. In the experiment about the sleep quality and use of phones, the fully randomized design will involve assigning subjects randomly according to a level of phone use. Use a random number generator to do this. When you use the block design, you can group all the subjects by age and then randomly assign them to the corresponding groups according to their phone use.

In our example temperature-soil respiration experiment, you randomly apply the warming procedures to certain soil plots, and then the number generator will choose the map coordinates related to a certain study area. To apply a block design, you can group the subject plots according to the average rainfall on them and then assign these plots to the appropriate groups.

There are situations when randomization cannot be used because it is impractical or unethical. So, a researcher can use certain non-random or partially random designs. The experimental design that does not use randomization is called quasi-experimental.

Comparison of Between-subject and Within-subject Designs

A between-subject design has some other names. They are an ANOVA design and an independent measure design. It means that individuals are assigned to only one level of experimental procedures. Or, as in social or medical experiments, there are matched pairs to be sure that each group has a set of subjects in equal proportions.

A within-subject design is also known as a repeated measure design. Every individual gets a corresponding procedure consequently and then you measure the responses to each procedure. These measures can relate to such an experimental design in which the effect appears over time and the responses of individuals are also measured over time.

A within-subject design can use counterbalancing or changing an order of procedures across the subjects. You need this format to be sure that the order of these procedures cannot affect the experiment results.

Let’s consider our examples. In the sleep and phone use experiment, a between-subject design means that the subjects receive the level of phone use randomly and this level is regarded throughout the whole experiment. A within-subject design means that the individuals refer to different levels, such as zero, low, and high use, in a row, and the order they are assigned to these levels can be randomized.

In the temperature and soil respiration experiment, a between-subject design implies that you apply warming to different parts of soil randomly and the soil in these parts is kept at the same temperature within the entire experiment. A within-subject design is applied in the way when every single part of the plot gets different temperatures, for example, 1, 5, 10, or more degrees, and these temperatures can change consecutively during the course of the experiment. The order they can receive different temperatures can be changed according to the objectives of every stage of the experiment, and it is random.

Measuring Dependent Variables

Here is the last step of the experimental research design. You have to choose the way for collecting data for your dependent variable. Such measurements should be valid and not contain errors or bias.

There are variables that can be measured objectively with special tools, for example, the temperature.

Others need to be processed to become measurable. In the experiment related to the use of phones and sleep, you can measure the dependent variable by asking participants to notify the time when they go to bed and get up regularly. Or you can suggest using a sleep tracker. All these measurements can be valid and provide the most reliable results for the further statistical analysis.

Final Thoughts

If you want to get the results that are relevant and valid, you need to develop such an experimental design that takes into account all the specific features of the system under research. It means that it is context-dependent.

If you know how to make up the experimental design that is the most appropriate for your conditions and subjects, your experiment will be successful, and you will be able to make the most valid conclusions.