Steps in Planning a Research Experiment
An important part of the design process is the verification that the device functions as expected. One can view this validation process as a parallel to hypothesis-driven research. Your design criteria specify the hypotheses that will be testing. For example, if you are designing a new sensor, one of your design criteria might be that it has a higher signal to noise ratio than a similar device that is already being marketed. You will then need to measure the signal to noise ratios for both sensors and compare them statistically.
Hypothesis testing involves the following steps:
1. State the hypothesis to be tested.
You will be looking for a hypothesis that can be tested statistically. Review the handout on statistical testing from senior seminar. One example would be to choose two cases of something and formulate the hypothesis that they are different. This naturally leads to a T-test. For example, “The blood pressure of patients with hypertension will be decreased by Altase (blood pressure medicine),” or “The blood pressure of patients with hypertension will decrease more by Altase than by Brand X (blood pressure medicine).”
Another hypothesis might be that one variable is correlated with another. For example, “Blood pressure is correlated with the number of cigarettes smoked per day.” In this case you would do a linear regression of the blood pressure vs number of cigarettes smoked and examine the p-value for this regression.
2. Formulate a context.
For example, if you are to test whether blood pressure correlates with number ofr cigarettes smoked per day, who will the subjects be? Will they be selected from an existing database or will a group of heart patients be selected and tested for blood pressure? Which hospitals will be used? Or, will you randomly select people off the streets, give them a questionnaire and then measure their blood pressure. On which street will you set up your experiment?
3. Formulate a theoretical model.
a. This will help you validate the experimental results.
b. By looking at the theoretical model, you will know what variables you must keep track of during the experiment. For example, if viscosity shows up explicitly in the model, you know you must measure it. Furthermore, you may need to measure some hidden variables, such as temperature, if some of the variables depend on it.
4. Design the experiment.
a. Define all variables to be measured
b. Sketch the experimental setup
c. State what equipment and supplies will be needed
d. State the method for data analysis
e. Describe the experimental protocol
5. Construct the experiment
6. Test the experimental apparatus
a. Perform any calibrations that are necessary
b. The best way to test the apparatus is to make sure that it provides the results you would expect from some simple experiment. For example, if you are looking at something related to pipe flow, you might look at pressure drops across a venturi, or pressure drop in the straight pipe (Poiseuille flow) and show that they correspond to theory.
7. Perform preliminary experiments
a. You should design your experiment to work the first time you try it, but do not be surprised if it does not. You need to try the apparatus out, which means following your experimental protocol from beginning to end. Odds are that you will find one or two flaws in your experiment that may need to be fixed. Also, you will gain experience in this stage that will allow you to run the experiment more quickly and accurately next time.
b. You will want to examine your results to make sure that they make sense. Even if it is clear that the results you have obtained are incorrect, run through the calculations needed to reduce the data and test the hypothesis. Create all graphs that are needed. If you do the analysis correctly, you will have set up the equations in Excel so that the next time you run the experiment the data analysis will be much easier.
8. Perform the experiment.
a. Make sure you record everything. This may include temperature, barometric pressure, time of day, and other variables that do not seem too important at the time.
b. It is a good idea to do some of your data analysis while you are performing the experiment to make sure that the data are in the right ball park. This will help you find errors in the experimental protocol and elsewhere.
9. Clean up the experiment.
10. Perform the data analysis, including statistical tests.
11. Write up the experiment and make conclusions.
© Steven A. Jones, 2004
Louisiana Tech University