Page 1 of 2 Statistics Assignment Choosing the Design of an Experiment Study Sheet Three Primary Principles of Experimental Design By now you probably know these by heart, but in case you don't here they are again. You must follow these principles when doing this Assignment: Replication is repeating an experiment on different units or under different conditions. Replication is first accomplished by designing and documenting an experiment so that it can be repeated. This reduces chance variation in the results and makes it easier to show that the results can be generalized to the population being studied (or even to other populations). Control is accounting for all other potentially influential variables (lurking variables), so you can be confident that any differences in the response variable between groups is due to the explanatory variable. In experiments this can be accomplished by randomization, matching, and setting up control groups that don't receive the treatment. For example, in a study that looks at the effects of a new medication, subjects can be randomly assigned to a control group that doesn't receive the treatment or that receives a placebo. Randomization is the practice of randomly assigning experimental units to treatment and control groups. This ensures that the effects of any lurking variables are cancelled out between groups. It allows you to be confident that differences in response between the two groups are due to the treatment. For example, in a study that looks at the effects of a new medication, subjects can be randomly assigned to a control group that doesn't receive the treatment or that receives a placebo. Three Common Experimental Designs You probably already have these burned into memory, too, but here they are again in case you need them. The most common is a completely randomized or randomized comparative design where subjects are randomly assigned to different conditions and the responses from those in different conditions are compared. For example, if we take 1,000 patients and randomly give 500 a new drug to control diabetes and then compare whether the symptoms are less severe for those who received the drug compared to the 500 who didn't, we're using a comparative design. The key is that we've randomly assigned the patients to different conditions. A randomized block design is similar, but also takes into account that the effects of an explanatory variable on a response variable may differ for subjects or units with different characteristics. In a block design, the subjects or units are first divided into blocks based on a characteristic, and then the subjects or units in the block are randomly assigned to the different experimental conditions. If you blocked on gender in the diabetes experiment above, you'd first separate the males and females, and then within each gender randomly assign subjects to receive or nor receive the drug. The comparisons of the severity of symptoms would occur within the blocks. For example, we'd compare men who received the drug with men who didn't receive the drug