Observational Studies & Experimentation

An observational study is a process in which data is collected WITHOUT imposing treatments on participants.

An experiment is a process in which data is collected by randomly assigning treatments to participants.

A factor is the independent variable, or the variable that the experimenter manipulates to create a response. The levels are the specific amounts/types of that factor. Treatments are the combinations of factors and levels.

The response variable is what the experimenter is recording.

Experimental units are who/what is receiving treatments...remember particiPLANTS?

Having at least one 15-minute warm-up session per week resulted in a drastic reduction in tears in the ACL. In a study involving 4500 adolescent girl soccer players in Sweden, one group was randomly assigned to warm up with a neuromuscular exercise session and the other was assigned no exercise session. The group with the neuromuscular session had 64% fewer ACL tears than the control group.

What is the factor(s)? How many levels are there? What are the treatments?

Factor: Warm-Up

# of Levels: 2

Treatments: Neuromuscular Exercise Session, No Warm-Up Session

Having at least one 15-minute warm-up session per week resulted in a drastic reduction in tears in the ACL. In a study involving 4500 adolescent girl soccer players in Sweden, one group was randomly assigned to warm up with a neuromuscular exercise session. This group had 64% fewer ACL tears than the control group.

What are the experimental units?

Experimental Units: 4500 adolescent girl soccer players

Having at least one 15-minute warm-up session per week resulted in a drastic reduction in tears in the ACL. In a study involving 4500 adolescent girl soccer players in Sweden, one group was randomly assigned to warm up with a neuromuscular exercise session. This group had 64% fewer ACL tears than the control group.

What is the response variable?

Response Variable: Number of torn ACLs

Is diet or exercise effective in combating insomnia? Some believe that cutting out desserts can help alleviate the problem, while others recommend exercise. Forty volunteers suffering from insomnia agreed to participate in a month-long test. Half were randomly assigned to a special no-dessert diet; the others continued desserts as usual. Half of the people in each of these groups were randomly assigned to an exercise program, while the others were assigned to not exercise. Those who ate no desserts and engaged in exercise showed the most improvement.

What is/are the factor(s)? How many levels are there? What are the treatments?

Factor(s): Diet & Exercise

# of Levels: Diet - 2 Exercise - 2

Treatments: No Desserts & Exercise, No Desserts & No Exercise, Desserts & Exercise, Desserts & No Exercise

Is diet or exercise effective in combating insomnia? Some believe that cutting out desserts can help alleviate the problem, while others recommend exercise. Forty volunteers suffering from insomnia agreed to participate in a month-long test. Half were randomly assigned to a special no-dessert diet; the others continued desserts as usual. Half of the people in each of these groups were randomly assigned to an exercise program, while the others were assigned to not exercise. Those who ate no desserts and engaged in exercise showed the most improvement.

What are the experimental units?

Experimental Units: 40 Volunteers

Is diet or exercise effective in combating insomnia? Some believe that cutting out desserts can help alleviate the problem, while others recommend exercise. Forty volunteers suffering from insomnia agreed to participate in a month-long test. Half were randomly assigned to a special no-dessert diet; the others continued desserts as usual. Half of the people in each of these groups were randomly assigned to an exercise program, while the others were assigned to not exercise. Those who ate no desserts and engaged in exercise showed the most improvement.

What is the response variable?

Response Variable: Improvement in Insomnia

There are two key groups in an experiment:

1. Those who could influence the results - experimental units, treatment administrators, etc.

2. Those who evaluate the results - judges, evaluators.

An experiment is single-blind when all the individuals in ONE of the groups are blinded. An experiment is double-blind when all the individuals in BOTH of the groups are blinded.

A placebo is a fake treatment that looks/feels/acts like a real treatment. ALWAYS use a placebo when possible.

Having at least one 15-minute warm-up session per week resulted in a drastic reduction in tears in the ACL. In a study involving 4500 adolescent girl soccer players in Sweden, one group was randomly assigned to warm up with a neuromuscular exercise session. This group had 64% fewer ACL tears than the control group.

Can this experiment be single-blinded? Double-blinded?

When the response variable involves opinion/judgement, it is always best to at least blind the evaluator.

The best experiments are always placebo-controlled, and double-blind.

1. Control

This is how we control KNOWN sources of variation! We must "even the playing field" across our groups so that the only difference between the groups is the treatment we administer.

2. Randomization

Random assignment of treatments is the ONLY tool we have against know/unknown bias.

3. Replication

We have to have more than one experimental unit in each group. The results of our experiment must happen multiple times (that's the replication part!)

1. Completely Randomized

Simple, gets the job done.

2. Block

Use blocking when you think there is something that makes groups similar and could affect the response variable! (age, sex, etc.)

3. Matched Pairs

More sophisticated block design. Match things up when you believe they are similar in most ways, so the only difference is the factor you control.

We always expect small differences between treatment groups, that's just life.

We say differences between treatment groups are statistically significant when differences are "large enough" and cannot reasonably be attributed to day-to-day variation