Principles of Experimental Design Control – controlling the effects of lurking variables Randomization – the use of chance to divide experimental units into groups Replication – repeating the experiment on many subjects to reduce chance variations in the results
Experimental Designs Statistical Significance – an observed effect too large to attribute plausibly to chance. Double blind experiment – the identity of the treatment is concealed from both the subject and the administrator
Good Morning America August 16, 2013 11:20-13:20 Experimental Designs Good Morning America August 16, 2013 11:20-13:20
Experimental Designs Correlation does not mean causation!
Experimental Designs Name of Group 1 ½ the sample Sample Group Name of Group 1 ½ the sample Name of Group 2 Compare the results IN CONTEXT Random Assignment
Experimental Designs Completely Randomized Design – all the experimental units are allocated at random among all the treatments. Block Design – the subjects or units are grouped by similarities that are expected to affect the response to the treatment. Random assignment of units to treatments is carried out separately within each block. (show graphic)
Experimental Designs Sample Group Name of Group 1 ½ the block Name of Block 1 Name of Group 1 ½ the block Name of Group 2 Compare the results IN CONTEXT Random Assignment Block Sample Group Block Name of Block 2 Name of Group 1 ½ the block Name of Group 2 Compare the results IN CONTEXT Random Assignment
Experimental Designs Matched Pairs Design is a common form of blocking which compares just two treatments. There are two types of matched pairs: Each block consists of 2 units as closely matched as possible. The units are assigned at random to the treatment. One subject gets both treatments, one after the other. Each subject is his or her own control. The order of the treatments is randomized.