DAWG has the objective of helping plan and analyze the most common statistical situations that students and researchers encounter in agricultural and biological research. These are ANOVA and regression experiments for normally distributed data. The purpose of this page is to assist you with identifying what statistical method/model best fits your experiment. This can be either for planning an experiment in advance, or for analyzing data you have already collected (the same steps apply).
Good statistical practice suggests that you choose your design before doing the experiment. So if you are in the planning stages of an experiment, go through the key below, and when you enter a specific DAWG module, use the provided example data to help fully understand how your experiment should be conducted, and what types of data need to be collected.
Another very important aspect of planning an experiment is deciding how large the experiment needs to be. This falls under the topic of statistical power. Once you have decided on ANOVA or Regression methods (see 1 below), then go to the Examples tab and choose ANOVA power or Regression power links to learn about software tools for sample size determination.
If you have data ready for analysis, go through the key below and you will be led to an appropriate DAWG module that will guide you in using SAS to analyze your data.
Experimental Design Selection Key
1. Do you want to test for differences among treatment means?
>> I know my ANOVA design, or
>> If you want help choosing your ANOVA design, go to red arrow 2 below.
No, I want to explore relationships among variables:
>> I know my Regression design, or
>> I want help choosing my Regression design (these pages under construction)
For ANOVA, you may choose among seven experimental designs via this key. The choice you make below will lead you to pages that enable you to then choose among seven treatment designs, with the choice to further refine the analysis with any combination of three specialized features (or no specialized features).
2. Are you blocking on a factor?
Yes >> Go to red arrow 3 below.
No >> You have a Completely Randomized Design (What is CRD?)
3. Is your block too small to contain all treatments?
Yes >> Go to red arrow 4 below.
No >> You have a Randomized Complete Block Design (What is RCBD?)
4. Does each block contain one and only one treatment?
Yes >> Go to red arrow 5 below.
No >> You have an Incomplete Block Design (What is Incomplete Block Design?)
5. At this point you should have two blocking factors, that we will call rows and columns. Why?
Are the number of levels for your row factor different from the
number of levels for your column factor?
Yes >> Go to red arrow 6 below.
No, they are equal >> You have a Latin Square Design (What is Latin Square?)
6. Do you have any other sources of variation besides rows and columns?
Yes >> Go to red arrow 7 below.
No, You have a Crossover Design (What is Crossover?)
7. Can your rows and columns be divided into squares (same number of rows and
columns), with each square differing from others due to a third blocking factor?
Yes >> You have an Multiple Square Latin Square Design (What is MS Latin Square?)
No >> Go to red arrow 8 below
8. You have a Switchback Design (What is Switchback?)
Last choice = Switchback: you should have rows and columns in a rectangular array, but you have an additional concern (as in red arrow 6) that row differences may differ for each column.