chapter 4 collecting data

Question 1

What is bias? When talking about bias, what should you always include?

Answer 1

Bias is when a sample leans one way from the truth about the population. Always talk about why there might be
bias AND if it is systematically an over or under approximation of the population.

Question 2

What is the difference between stratified and cluster samples?

Answer 2

Stratified sampling creates groups that have similar context within and different between. You pick some subjects from each group so each group is represented.

Cluster sampling has heterogenous groups. Randomly select some groups and then survey everyone within the group.

Question 3

Why do we use random selection?

Answer 3

Random selection allows us to generalize the results of context to the target population context.

Question 4

What is the difference between stratified and blocking?

Answer 4

Strata is used for sampling, right before you SELECT the experimental units. Blocking is used for experiments right before ASSIGNING treatments to the experimental units.

Question 5

What is the benefit of random assignment?

Answer 5

Random assignment helps us balance the confounding variables (context) and allows us to say context CAUSES context. Note: whenever you say the words confounding variable you always must state an example of one.

Question 6

When can we say causation vs. association?

Answer 6

We can say causation when treatments are applied or random assignment is used, (in an experiment!). If no treatments were applied, or random assignment is not used, then we can only comment on an association. Note: we do NOT say correlation!

Question 7

How do you choose strata (in sampling) or blocks/matched pairs (in experiments)?

Answer 7

You want similarity context within each strata or block AND differences context between the strata or blocks/pairs.

Question 8

What is a statistical advantage of using blocking or matched pairs compared to a completely randomized design?

Answer 8

A blocking or matched pairs design will have less variability compared to a completely randomized design which will allow the results to be more precise (NOT ACCURATE).
Therefore, when concluding we will be better able to distinguish the effects of the treatments (context) on the response variable (context).

Question 9

How do you select a SRS? (This could be used for selecting a random sample, OR for creating random assignment in
an experiment.)

Answer 9

1. Label ________ with a number (specify!)
2. Use a RNG from _1___to __n__to select n numbers –address repeats!
3. Connect the numbers selected back to the subjects/things that were labeled to use as a sample, or to create a
   treatment group. (If using this process for random selection, the final group can be created by just stating “all
   others will be in treatment group _______.”)

Question 10

How do you randomly assign treatments to matched pairs?

Answer 10

For each matched pair of experimental units, label one unit as A and label the other unit as B. For each pair of units, toss a coin. If the coin lands on heads,
unit A gets the first treatment context and unit B gets the other treatment context. If the coin lands on tails, unit A gets the 2nd treatment context and
a gets the first treatment context. Repeat for all pairs.
OR
Label the members of each pair of units as “1” and “ 2.” Using a random number generator, generate 1 integer from 1 to 2. Give the first treatment context to the unit whose number is selected and the other treatment context to the unit whose number is not selected. Repeat for all pairs of units.

Question 11

What are some key differences between an observational study vs. experiment?

Answer 11

An experiment has treatments that were randomly assigned – an observational study does not. As a result, an
experiment can show causation – an observational study can only show an association.

Question 12

What are the steps to explain a confounding variable?

Answer 12

1. Connect the confounding variable to the explanatory variable.
2. Connect the confounding variable to the response variable.
3. So, we don’t know if it is the original explanatory variable or the confounding variable that causes ___________.
   Example: Explain a confounding variable for an observational study on the impact of taking on SAT prep course on a
   person’s SAT score.
   People who take an SAT prep course could be more motivated to do well on the SAT test, and that motivation could
   lead to higher SAT scores. So, we don’t know if it is the prep course or the higher motivation that causes higher SAT
   scores.