Experimental Designs MATCHED PAIR DESIGN Two subjects with

Experimental Designs

MATCHED PAIR DESIGN �Two subjects with similar attributes are deliberately paired. �The treatment being tested is randomized.

You want to test two pair of running shoes. It just so happens that your track star friend’s favorite pair is one of them (Brand X). The other is a new brand (Brand Y). q You flip a coin. If it lands on heads, your friend wears Brand X first and you will time her as she goes around the track; if it lands on tails, she wears Brand Y. You time her. q After she has gone around the track once with the first pair of shoes, she goes around it again, but with the other pair of shoes this time. You time her again. q

RANDOMIZED COMPARATIVE DESIGN � Compares two or more treatments. � One explanatory variable. � The experiment can be stopped early if it is clear that the treatment has experienced much more success than the placebo (STATISTICALLY SIGNIFICANT) or if it has determined that the treatment is dangerous.

• The patients may have been divided by randomly assigning odd/even numbers to each patient. • This may be the reason for the unequal number in each group. • There also other ways of random assignment… can you think of any? ? ?

DID YOU THINK OF THESE? ? ? � � You can use numbers: 0 -4: 1 st treatment 5 -9: 2 nd treatment You can use coins: Heads: 1 st treatment Tails: 2 nd treatment You can’t use gender or age in a randomized comparative experiment, unless… � You first make it a block design

WHAT’S A BLOCK DESIGN? ? ? • • A block design is a group of subjects that have some commonality (such as gender, age, etc. ) that is known before the experiment, and that could possibly affect the response to the treatment. This creates two or more individual randomized comparative experiments.

1. 2. In this example, men and women are separated, and then randomly assigned into 3 groups in order to judge their reactions to 3 advertisements. The purpose of this experiment is to see if men react differently to these advertisements than women do.

RANDOMIZED COMPARATIVE DESIGN BLOCK DESIGN Compares one explanatory variable � Random assignment of subjects for treatments � Not able to divide subjects into subgroups � Can compare more than two treatments at once � � Compares one explanatory variable � Random assignment of subjects for treatments � Able to divide subjects into subgroups � Can compare more than two treatments at once

QUESTION… Wouldn’t it just be easier to test the product and nothing else? Like if you were to test a new and improved Tylenol and only that one medicine. Then you would have results and you would only need half the subjects. Simple, right? � This is called a “One Track Design” � There is nothing to compare the subjects’ reactions to, so there could be many chances for lurking variables to occur. � Therefore this would not be an effective experiment. �

COMPLETELY RANDOMIZED DESIGN � All subjects are randomly assigned to groups and all groups are given different treatments. � This design can have any number of explanatory variables, not just one.

SUCH AS… � This experiment has two explanatory variables: temperature and cleansing agent. A Completely Randomized design can have two, or even more than two, explanatory variables. � You do want to be careful not to have too many explanatory variables, as it will also defeat the purpose of the experiment. Figuring out the best treatment is warm water with regular tide (Treatment #4) will cover many people; adding in with a particular brand model washing machine and specific water level may lead to confounding.

� � � You can’t pick and choose who receives which treatment. You need to have two or more treatments so lurking variables are minimized. You need to include enough subjects to reduce the chance of variation. � � More subjects = better chance of producing groups that match closely = less variation. If you do all of this correctly, any differences are because of the effects of the treatment.

1. 2. GROUP WORK #11: TEXTING 3. Do you think you could control the number of texts you sent in a month if your plan set a limit before it started charging you outrageously? Using 3 groups: 1. Phone company will send you a text when you have 5 texts remaining. 2. Phone company will give you a chart and information about monitoring your texting use. 3. Phone company will give you the number of texts you are allowed in a month, but no help in monitoring or limiting them. (Control Group) Design two different experimental designs.