STAT 104 - Chapter 9: Experiments & Chapter 10: Data Ethics

  1. 1. What is the difference between an observational study and an experiment?

    2. What is the purpose of an experiment?
    1. An observational study merely observes individuals. They may be measured or asked questions, but no attempt is made to change the reponse or influence behaviour.

    *Causation cannot be inferred because of the possibility of lurking variables*

    In an experiment on the other hand, individuals are randomly assigned to various treatments in order to observe their responses. They do not choose which treatment they will receive. 

    2. The purpose of an experiment is to establish whether the different treatments cause the response to change* (causation)
  2. 1. How can we tell if a study is actually an experiment or simply an observational study?

    2. What are confounding variables?

    3. Give examples of confounding variables in the excercise on the relationship between mortality and the intake of red meat.

    Explain why they are confounded
    1. The subjects must be randomly assigned to groups. If they are not, it is an observational study. 

    2. Two variables (explanatory or lurking) are said to be confounded if their effects on a response variable cannot be distinguished from one another. 

    3. Variables that are confounded with the intake of red meat are: alcohol consumption, physical activity, BMI and smoking.

    Because people who eat red meat are also more likely to drink alcohol, smoke, excercise less and have a higher BMI, we cannot establish causation.
  3. Define the following terms:

    Levels of a factor
    Subjects: The individuals of interest are often called subjects in an experiment, especially if they are people

    Factors: The explanatory variables in an experiment are called factors.

    • Levels of a factor: The experimenter chooses the levels of each factor.
    • ex. Diet pill and placebo

    Treatments: The treatments are the conditions experienced by the subjects during the experiment.

    Response: The response variable is the variable that is measured at the end of the experiment.
  4. 1. What types of variables can we use as factors in an experiment, and what types can't we use?

    2a. What do you write when describing treatments if there is only one factor?

    b. What do you write when there are two or more factors? (put on cheat sheet)
    1. In an experiment, the factors must be variables that can be manipulated (ie. changed) by the experimenter.

    Pre-existing variables (eg. gender, age, daily temperature) cannot be changed by the experimenter so they cannot be used as factors.

    2a. If there is only one factor, the treatments are the levels of the factor.

    b. If there are two or more factors, then the treatments are all combinations of levels of the factors.
  5. Image Upload 1
    Identify the following:

    Levels of the factors:

    2. How would you draw a diagram for this completely randomized experiment? (3 point question on exam)
    Subjects: 50 adolescents 14 to 18 years old with BMI greater than 25 (obese)

    Response: Weight loss after 2 years 

    Factors: Type of intervention (this is what they were manipulating)

    • Levels of the factors: 
    • 1. Gastric banding
    • 2. Lifestyle intervention

    Treatments: Gastric banding and lifestyle intervention (because there is only one factor)

    • 2. Diagram of completely randomized experiment
    • Image Upload 2
  6. 1. What is a completely randomized experiment

    2. What are 2 ways randomization can be done for an experiment?
    1. Subjects are randomly assigned to the various treatments.

    If the number of subjects in each group is the same, then the experiment is called "balanced."

    If the number of subjects is not the same, it is "unbalanced," but this is not a problem because the computer software will handle the numbers. 

    2a. Random number table method: Give each of the subjects a number of the same length (eg. if there are 180, label them 001, 002, 003...180). 

    If there are 3 groups, use the random number table to select 60 numbers for the first group, do the same for the second group and the remaining will go in the third group. 

    2b. Slips of paper: Say we have 180 subjects and three treatments. Cut up 60 slips of red paper, 60 green and 60 blue. Mix the slips up in a box. Without looking at the box, hand each subject a slip of paper. 

    *Remember, they are not allowed to choose the slips themselves.
  7. Describe the three principles every good experiment should include.
    1. Control the effects of lurking variables on the response by having more than one treatment

    2. Randomize to equalize the chance that lurking variables will affect all treament groups.

    If the subjects are not randomly assigned, then lurking variables may be more likely to affect particular groups.

    3. Replicate. Each treatment group must have more than one subject. 

    If there is only one subject in each group, then we cannot tell whether the treatment or some lurking variable is causing the response to change. 
  8. 1. What is a controlled experiment compared to an uncontrolled experiment?

    2. What are 2 common misconception about controlled and uncontrolled experiments?

    3. What is a control group anyways?
    1. A controlled experiment has more than one treatment, whereas an uncontrolled experiment only has one treatment. 

    2a. An uncontrolled experiment is thought to be an experiment that has no "control" group. This is not the case. An uncontrolled experiment only has one treatment.

    2b. Another widely misunderstood idea is that every experiment must have a control group. This is not correct. It is not necessary for an experiment to have a control group, and most don't.

    ex. Adolescent obesity, wine, beer or spirits etc. 

    3. The control group is defined as the group in an experiment or study that does not receive treatment by the researchers.

    It is then used as a benchmark to measure how the other tested subjects do.

    ex. Zoning out. One group received alcohol and were told to read War and Peace, and one group recieved what they thought was alcohol, but wasn't. This was the control group.
  9. 1. If you don't have ______, then you can't consider it an experiment. 

    2. What is a double blind experiment, and what is a single blind experiment?

    3. Give an example of an experiment that suffers a lack of realism.
    1. More than one treatment group. Every experiment must have more than one treatment group. 

    2. Double blind: If neither the experimenter nor the subjects know which treatment each subject is receiving. Treatments may be labeled A, B, C, etc. so that they cannot be identified. 

    Very often, this is just not possible.

    Single blind: The subjects do not know which treatment they are receiving, but the experimenter does know.

    The subjects know only that they are receiving treatment A, but cannot identify useful information about treatment A. 

    3. Does having more babies lead to lower stress levels in humans? An experiment randomly assigned female rats to either be allowed to breed or not allowed to breed. The rats that were allowed to breed showed lower levels of stress than those that were not allowed to breed.

    This experiment suffers from a lack of realism because stress is a very complicated variable, and therefore we cannot generalize from rats to humans. 

    We don't know how stress affects rats and humans differently.
  10. Mangoes are generally harvested at the mature green stage and ripen up during the marketing process of transport, storage, and so on. 

    During this process, about 30% of the fruit is wasted. Because of this, the impact of harvest stage and storage conditions on the postharvest quality are of interest.

    In this experiment, the fruit was harvested at 80, 95, or 110 days after the fruit setting (the transition from flower to fruit) and then stored at temperatures of 20°C, 30°C and 40°C. 

    For each harvest time and storage temperature, a random sample of mangoes was selected, and the time to ripening was measured. 

    1. Assume that 90 mango flowers have been selected for this experiment. Identify the following:

    Levels of the factors:

    2. How would you draw the diagram for this completely randomized experiment?
    1. Subjects: 90 mango flowers

    Factors: 2 factors: Harvest time and storage temperature

    Levels of the factors: Harvest time has 3 levels: 80, 95 and 110 days

    Storage temperature has 3 factors: 20°C, 30°C and 40°C.

    • Treatments: 
    • 1. 80 days and 20°C
    • 2. 80 days and 30°C
    • 3. 80 days and 40°C

    • 4. 90 days and 20°C
    • 5. 90 days and 30°C
    • 6. 90 days and 40°C

    • 7. 110 days and 20°C
    • 8. 110 days and 30°C
    • 9. 110 days and 40°C

    Response: Time to ripening

    • 2. Diagram
    • Image Upload 3
  11. What are the 5 things you need to identify when setting up a completely randomized experiment?
    • 1. Subjects:
    • 2. Factors:
    • 3. Levels of the factors:
    • 4. Treatments:
    • 5. Response:
Card Set
STAT 104 - Chapter 9: Experiments & Chapter 10: Data Ethics
Final Exam Prep