Please enable JavaScript.

Coggle requires JavaScript to display documents.

Unit 4: Producing Data (Samples and Surveys (Bias (Definition (Bias is…

- - - - Lurking Variable
        
        For example, if psychologists were studying the relationship between the reading abilities of children and average time they spent reading every day, their results could be affected by the natural ability of the children.
        
        A variable that is not among the explanatory or response variables in a study but that may influence the response variable
  - - - Factors - conditions being changed in order to look for a response (explanatory variable)
      - Treatment - a specific condition applied to the individuals in an experiment
      - Experimental Units (subjects) - the smallest collection of individuals to which treatments are applied. People are called subjects.
      - Response Variables - what is being measured (difference between initial and final observations)
      - Three Principles of Experimental Designs
        
        Control: isolate one experimental group that will not receive any treatment, and ensure that the only systematic difference between all the groups is the treatment administered.
        
        Random assignment: use impersonal chance to assign treatments to different experimental groups, so that there are roughly equivalent groups of experimental units and the effects of lurking variables that are not controlled can be balanced.
        
        Replication: Enough experimental units are used in each group so the experiment can be repeated and similar conclusions can be drawn. This reduced the chance of variation by averaging the observations,
      - Placebo Effect and Blinding
        
        Placebo effect - the placebo effect is when people believe that a treatment is going to make them feel better, so they feel better.
        
        Blinding
        
        Single-blind: Typically, the subject does not known the treatment they are receiving and those measuring do. It is when one party knows the treatment being measured and the other one does not
        
        Double-blind - if both of the subjects and the those measuring the response variables, don't known what treatment the subject is receiving
      - Experimental Design
        
        In order to prevent bias as a result of the placebo effect, we are going to design the experiment with a single-blind procedure. Obviously, the students are going to be aware of which breakfast they ate. But the administrators of the test will be unaware of which subject received which treatment. The test administrators might believe that one breakfast will produce better results than another breakfast, and grade differently based on the results that they expect to see.
        
        Summary: The experiment will be measuring the test scores of students vs the quality of their breakfast. 15 students will be randomly selected from one school by using a random number table. These 15 students will be given the same breakfast, and then take a 10-question reading exam. The next day, they will be randomly selected again and separated into three groups of 5, each group receiving a different breakfast. One group will eat no breakfast, the second group will have only a small bowl of sugary cereal, and the third group will eat a protein-filled breakfast sandwich. The students will then take the same 10-question reading exam again. The score differences from the original test and the second test will be compared to see the effect of breakfast on test-taking abilities.
        
        Design Map
- - - - Definition: Choosing individuals who are easiest to reach
      - Bias: Generally over or under predicts the value since using data from one small location does not accurately predict for a large location
      - Example: A survey of one's Instagram followers would be simple, because they are all in the same place. However, this would lead to bias, as one's followers are typically a part of the same demographic.
    - - Definition: Consists of people who choose themselves by responding to a general appeal
      - Bias: People who have strong opinions on the topic are more likely to respond, so it often shows data in the same direction
      - Example: An optional survey about school policy that is emailed to all students. The students that respond are likely the ones that are most passionate about school policy, and thus might be more likely to be in support.
    - - Definition: When some individuals are not included in the data since they are left out of the process of choosing the sample
      - Bias: People are left out of the sample
      - Example: A list of students taken from the parking lot registration. This would be an example of under-coverage, because this population does not include the students who either get a ride to school or take the bus.
  - - - Definition: When an individual in the sample does not participate in the survey
      - Example: If a survey is conducted on how many hours per week the an adult in Massachusetts sleeps on a sample of 10,000 randomly selected but 2,000 of the individuals do not respond, they data would not accurately predict the number of values
    - - Definition: Confusing or misleading questions can drastically change the outcome of the data
      - Example: A survey question about whether a particular children's toy store should be shut down asks, "Do you want your children to live in a world without toys?" This wording would lead the surveyed people to the obvious answer, "No", even if they do not support this particular restaurant.
    - - Definition: when people respond incorrectly to the questions
      - Example: A survey question has two answers, which vary only in the fact that one says "would" and the other says "wouldn't". The individual taking the survey moves too quickly and cannot see the difference, so just chooses one of the choices.
  - - - Bias is seen in statistical studies where certain outcomes are systematically favored.
    - - Bias in samples and surveys can be caused by the sampling and non-sampling errors, which are listed above.
    - - Bias in samples and surveys can cause the sample to either overpredict or underpredict the actual population. Surveys should be representative.
  - - - Advantages: more representative of the population's different perspectives than simple random sampling, takes variation between different people into account
      - Disadvantages: Finding the characteristics to separate the population by can be difficult, difficult to put people into specific boxes, as a result of difficulty can be very time consuming
      - Definition: the population of individuals is classified into groups of similar individuals, then a separate SRS is chosen in each stratum and all SRSs are combined to form the full sample
    - - Advantages: more convenient than stratified SRS, different within each cluster
      - Disadvantages: small clusters may not represent whole population (ex. if clusters are formed by homerooms that are separated by grade and more senior homerooms are selected than freshman it would not accurately represent the entire school), convenience sampling error
      - Definition: when the population is divided into smaller groups that mirror the characteristics of the population, then clusters are chosen randomly to be surveyed
    - - Advantages: doesn't favor any individuals, works well for small samples, uses tables of random digits to select data randomly
      - Disadvantages: time-consuming with large samples or if the individuals are spread out over a large area, doesn't evenly represent individuals (more males could be selected than females)
      - Definition: all individuals in the sample have an equal chance of being selected