Module 3

Question 1

What is QuODDID stand for?

Answer 1

Qu : Question of scientific interest

O; Outcome measures

D: Design

D: Data analysis

I: Interpretation

D: Dissemination

Question 2

What is the manual of operations?

Answer 2

A protocol/statistical analysis plan in which you write down exactly what you plan to do before beginning study. Document when and why you deviated from something

Question 3

Secondary questions involve _____(1) and can help paint a more complete picture of what’s going on.

Answer 3

(1) stratified subgroup analysis

Question 4

What are the three desirable characteristics of a research question?

Answer 4

1. Relevant: Is the question related to the topic of interest?
2. Precise: Is the population of interest specified? Is relationship of interest specified?
3. Testable by the study: Can we actually assess and answer the question within our study

Remember: Can you Really Print That?
location: brody

Question 5

A fuzzy research question can live one open to critiques about:

Answer 5

1. Research methods not matching th question you intended to ask
2. Methods selected to try and increase your chance of a significant finding after the fact

Question 6

What are the three desirable characteristics of outcome measures?

Answer 6

1. Relevant
2. Precise and accurate
3. “Movable” : likely to be influenced by the interventions

Question 7

Our ability to achieve accuracy and precision is impacted by _____(1).

Answer 7

(1) the level of granularity we use to measure things

Question 8

What is the difference between accuracy and precision?

Answer 8

Accuracy: All dots near the centre of the bullseye, but they are all not concentrated in one area.

Precision: All dots concentrated in one area but not near the centre of the bullseye

Question 9

When we talk about precision and accuracy, what are the three methods we can use? Order them from best to worst.

Answer 9

Numerical measurement > ordered categories > binary categories

If we wanted to measure how fast someone could run:
1. Running speed
2. Is the subject fast/moderate/slow
3. Is the subject fast or not fast

Question 10

What are the two types of study designs?

Answer 10

1. Experiment: Assignment to treatment groups is controlled by the investigator
   - treatment assignment my be done using a randomized mechanism or not
   - treatment may be masked/blinded (single, double, triple) or not

2, Observational study: Assignment to treatment groups is not controlled by investigator. Participants usually self-select

Question 11

Who are the four stakeholders in experimental design?

Answer 11

1. Investigator: Individual asking the question in design and designing the study to answer it
2. participant: individual being observed to understand the impact of the intervention on the outcome measure
3. Assessor: Individual evaluating the outcome measure for each of the participants
4. Data analyst: Individual analyzing the data

Question 12

What is single, double, or triple blinded?

Answer 12

Single: The participant doesn’t know which treatment group they are in. Assessor and data analysis know.

Double: Participant and assessor of outcome dont know which treatment group the participant was assigned to. Only the data analyst knows. This protects against assessor evaluating measures in a biased way

Triple: Assessor, participant, and data analyst dont know about treatment groups. This prevents bias from data analyst.

Question 13

What are the two subcategories of experimental study design?

Answer 13

1. Parallel group study: Subjects allocated to receive only one “level” of the treatment being compared. Comparison of outcomes is done ACROSS experimental units(people). Make comparison groups similar through randomization
2. Cross-over study: Each subject gets multiple “levels” of the treatment. Comparison of outcomes is within experimental units(people). Each unit acts as “own control/comparator”. Randomization here occurs in terms of what treatment the groups receive first.

Question 14

In a cross-over study design, you no longer need to worry about comparing groups to groups. Comparison is within each participant. This minimizes _____(1) and there is thus less of a chance of _____(2).

Answer 14

(1) the external characteristics that can influence the primary relationship we care about

(2) confounders

Question 15

What is the challenge with cross-over studies?

Answer 15

WASH OUT

How do we ensure that the impact of the first exposure/intervention doesn’t linger and affect the outcome measure of their second exposure/intervention?

Question 16

What two outcome measures did we use for the chocolate trial?

Answer 16

1. Individual ratings for two chcolates
2. Direct comparison of two chocolars

Question 17

What is a DAG?

Answer 17

Directed acyclic graph the illustrates relationships between various variables

Question 18

In a DAG, variables in rectangles are ___(1) while variables in ovals are ____(2)

Answer 18

(1) oberved/measured
(2) not

Question 19

In a DAG, lines(arrows) represents a _____(1) between variables.

Answer 19

(1) causal relationship

Question 20

Absence of lines in a DAG means _____(1) given the other vairbales

Answer 20

conditional independence

Question 21

Randomization can ____(1) from observed and measured variables, from unobserved and unmeasured variables, and even unknown variables

Answer 21

(1) mitigate confounding varibales

Question 22

What is the importance of removing confounding variables in an experiment?

Answer 22

Allows us to be sure that the associations we observe are reflective of the true relationship of the two variables we want to study

Question 23

In our chocolate trial, what were possible confounders and how did we remove them from out experiment?

Answer 23

1. Tasting order
2. Chocoholic gene

Question 24

When the outcome variable of interest is _______(1) instead of continuous, we talk about ______(2) instead of means

Answer 24

(1) dichotomous(binary)

(2) proportions

Question 25

How do you calculate the confidence interval for a single proportion?

Answer 25

Refer to module 3 cheatsheet in notes just know how to calculate p hat, and be familiar with all the terms in the formula for SE

Question 26

How do you do a hypothesis test for a single proportion?

Answer 26

Calculate confidence interval as you would for a single proportion and then determine the all value and see if it is contained in the CI. If it is, then we fail to reject the null hypothesis

Question 27

The CI for a difference in proportions is used for questions that seek to compare groups and populations based on their proportions. How do you calculate the CI for a difference in proportions?

Answer 27

Refer to module 3 cheatsheet

Question 28

How do you do a hypothesis test for a difference in proportions?

Answer 28

Determine 95% CI for a difference in proportions and see if null value is contained within that interval.

Question 29

What is the three step process involved with hypothesis testing?

Answer 29

1. Consider your research question 2. Define null hypothesis 3. Test you null using your observed data

Question 30

What is the Texas sharpshooter fallacy?

Answer 30

The name epidemiologists have given tot the tendency to assign unwarranted significance to random data by viewing it post hoc in an unduly narrow context.

Question 31

What is p hacking?

Answer 31

the misuse of data analysis to find patterns in data that can be presented as statistically significant

Question 32

What is the key idea behind multiplicity?

Answer 32

Events that are rare on an individual level are still likely to happen across a large group of individuals

Question 33

A type 1 error is rare on an individual hypothesis test, but ____(1) to happen at least once across a large group of tests.

Answer 33

likeyl

Question 34

Compare hypothesis generating vs hypothesis testing.

Answer 34

Hypothesis generating: Lots of questions, lots of tests Hypothesis testing: few questions, few tests

Question 35

What are the bonferroni corrections?

Answer 35

1. divide p-value by the number of comaprisons 2. Or make a confdience interval thats like (100 - 5/(number of comparison)

Question 36

How would you do multiplicity problems

Answer 36

calculate the probability of the event happening independently and then apply a binomial distribution using that p value

Question 37

What is a confounding variable? Draw a diagram, and identify confounding, predictor, and outcome variables.

Answer 37

Look at diagram in the notes sheet Factor Z would be a confounder if - They differ between levels/categories of X AND - They have a causal relationship with Y

Question 38

What does it mean if something is stratified?

Answer 38

Creating subsets of the data and looking at data on those groups

Question 39

What is Simpsons paradox?

Answer 39

Simpson's Paradox refers to a phenomenon in which a trend appears in different groups of data but disappears or reverses when these groups are combined. In other words, the overall percentages in two groups (the treatment and control groups) can be misleading because of a confounder. Berkeley's admission data is an example of Simpsons paradox

Question 40

What are two ways we control for confounders?

Answer 40

1. "Control for" potential confounding variables: conduct out comparison of interest within confounding groups, like within a department fro the Berkeley admissions data or within a poverty group for our MSCD and medical expenditure question 2. Design an experiment where the treatment and control groups are "otherwise similar" by design.

Question 41

What is effect modification?

Answer 41

Effect modification occurs when the effect of a single exposure on an outcome depends on the values of another variable ex. in the chocolate trial it would be called effect modification if the effect of cocoa content on preference depends on a thrid variable

Question 42

Why is stratifying useful?

Answer 42

Reduce variance rather than bias answer a secondary question of interest - stratify to look at difference in preference by third variable