designing your experiment 

Designing an experiment is the fundamental tool for collecting valid data. Data is important because it either supports or does not support your hypothesis. It really doesn't matter which of these results because you learn something about your topic and can report important findings despite the outcome if you follow experimental protocols. Look at the following graphic. It shows the close connection between the testable question, your hypothesis and your experimental design. It is essential that you design an experiment that will generate data related to your testable question and hypothesis. There are details below that describe how to turn your testable question into a hypothesis and how to design an experiment that will test your hypothesis.

• Step 1: Understanding Variables
  • There are 2 types of experimental variables. They can be confusing; however, it is essential that you understand them. The variables are called the independent and dependent variables. Each of them is defined below and then you are given several examples.
    • The independent variable is the variable that you manipulate or change.
    • The dependent variable is the variable that is affected by the changes you make on the independent variable. This is the variable that you will measure.
• Step 2: Forming a Hypothesis
  • Use your testable question as a starting point. Here is the testable question: Will the wavelength of light change the rate of photosynthesis?
  • Then develop your hypothesis using the IF/THEN format.
    • The IF/THEN format is a stress-free method for writing a hypothesis. The IF/THEN format allows to describe a causal relationship. Causality means that a set of conditions or events cause something else to change. The mission of experimental research is to illuminate this causality. The format allows you to make a statement as follows: IF this is true, THEN I expect this to happen. The format detail including the variables follows: If (the independent variable) does this, then (the dependent variable) will result. For those of you who relate to your mathematics training, the independent variable in a hypothesis is found on the x-axis, while the dependent variable is found on the y-axis.
    • For example, there are 2 VARIABLES in this hypothesis statement; the independent variable (ex. wavelengths of light: the variable you manipulate/change) and dependent variable (ex. the rate of photosynthesis: the variable that is affected by the manipulation or the variable that you will measure).
  • Examples of hypotheses:
    • IF the rate of photosynthesis is related to wavelength of light, THEN exposing a plant to different colors of light will produce different amounts of oxygen.
    • IF the rate of photosynthesis is related the temperature of the environment, THEN exposing a plant to different temperatures will produce different amounts of oxygen
    • IF the rate of photosynthesis is related to the surface area of the leaf, THEN differing sizes of leaves will produce different amounts of oxygen.
• Step 3: Designing the Experiment
  • You will need to know about the following before you can design your experiment.
    • Controls. Do not be confused with the term "controlling the variables" when you begin to design your experiment. This should not be confused with your independent and dependent variables. This terminology "controlling the variables" describes the effort to control your experimental environment as much as possible so that you are limiting the influence of other factors. For example, you can think of limiting the variables as limiting outside influences. Outside influences are anything that can affect the outcome of your experiment other than the independent variable. Let's use the independent variable of wavelengths of light in this hypothesis statement as an example; IF the rate of photosynthesis is related to wavelength of light, THEN exposing a plant to different colors of light will produce different amounts of oxygen. If you allow other unmeasured light sources to contaminate your experiment, then you cannot make any conclusions about the wavelengths of light. This is like a person contaminating a crime scene. The "evidence" is not valid because the scene was contaminated.
    • Data collection. The purpose of an experiment is to collect data. At this level, you should attempt to collect as much data as possible. This means that 100 to 1000's of data points are something you should aim toward. This can be accomplished if you collect data on the same set of subjects a number of times. Below are some examples.
      1. You have 7 chameleons and you want to test their ability to climb across Teflon surfaces. You should conduct 20-30 trials with each chameleon for each surface you test. This gives you 140-210 sets of data for each chameleon for each surface. This gives you the confidence to make some conclusions based upon your data.
      2. You want to test the force needed to keep ice from freezing. After you design a method to measure this force, you will need to conduct 50-100 tests. Why? Because many trials will allow you to generalize your results.
      3. Too few trials may give you an inaccurate picture of your work. Think about this scenario. If your class had a test and you knew the test scores of 2 students; 1 with a 100% and one with a zero, you may assume that the class had a 50% average. However, when you add the test scores of all of the 30 students in class adding more data to the set, you would find out that the class average was 85%. This is a more accurate picture of the test because it is based upon more data.
    • Design the experiment. Now you can begin the actual design process for collecting your data. There are several kinds of designs you can use. The choice of design will also guide the type of data you collect and the statistical measures you use to examine your data. Making the correct choice will allow you to ease into the next phase of this project. These are some general categories of research design; experimental, quasi-experimental and non-experimental. Each of the designs is described below.
      • Traditional Experimental Research Design. This design is utilized when a researcher tries to control all of the variables possible.
      • Quasi-Experimental Research Design. This design is utilized when nature controls the variables while the researcher selects what they will measure (ex. crime rate, moon phases).
      • Time Series Experimental Research Design. This design is utilized for before and after studies (ex. before and after study).
      • Non-Experimental Design. This design involves no hypothesis testing. Therefore, it is not the best choice for your HS3 project.

• Ask yourself these questions:
  1. How much control can I exert over the variables I want to study?
     • Answer: A lot: Decision; Choose the traditional experimental design
     • Answer: Very little; Decision: Choose the quasi-experimental design
  2. Where will my data come from?
     • Answer: I'll generate the data; Decision: Choose traditional time series
     • Answer: I'll use existing data; Decision: Choose quasi-experimental design
  3. Am I as interested in how something happens as what happens? Will my experiment occur in phases, continuously monitoring variables over time? Will I be working in a single group rather than several?
     • Answer: Yes; Decision: Choose time series
     • Answer: No; Decision: Choose traditional or quasi-experimental
• This section will further describe each category of experimental design.
• Traditional Experimental Research Design.
  • The experimenter controls and manipulates the variables. Some examples are listed below in hypothesis format.
    • If the level of carbon dioxide increases in a closed environment; then the rate of growth of grass will be reduced.
    • If oil is used on 90 degree surfaces; then chameleon "stick time" will be reduced.
• Quasi-Experimental Research Design.
  • The experimenter can't control or manipulate the variables; however, they can test a hypothesis.Some examples are listed below in hypothesis format.
    • If the atmospheric pollen count is above 61 per cubic meters of air; then high school absenteeism is high.
    • If students take at least 2 AP classes, then they are more likely to graduate college within 5 years.
• Time Series Experimental Research Design.
  • The experimenter may manipulate variables, observes changes over a period of time and tests a hypothesis. Some examples are listed below in hypothesis format.
    • If males and females exercise equally; then male and female strength as tested by a force meter will increase equally.
    • If several samples of iron are treated with increasing levels of WD-40, then the rate of rust accumulation will decrease.