Sample Size

Question 1

What is a data codebook?

Answer 1

Mapping of all variables, values, etc.

Question 2

How can we determine sample size?

Answer 2

1) Power analysis

2) Precision estimation

Question 3

What is statistical power?

Answer 3

The probability of rejecting the null when it is false.

Question 4

What is precision estimation?

Answer 4

Sample size using inversion of a confidence interval.

Question 5

What is MCAR?

Answer 5

Missing completely at random data. (Ex: car breaks down).

Can ignore or delete.

Question 6

What is MAR?

Answer 6

Missing at random.
Missing data may depend on observed data (older PX miss follow up).
Control for the variable in the data.

Question 7

What is NMAR?

Answer 7

Not missing at random.
Depends on unobserved data (Ex: SBP is more likely missing for hypertensive PXs).
Cannot ignore this data.
We will try to avoid having it in study.

Question 8

What is the gold standard for handling missing data?

Answer 8

Multiple imputation.

Question 9

Describe multiple imputation.

Answer 9

Missing values are replaced by several simulated data sets.

Question 10

Why is missing data an ethical issue?

Answer 10

1. Reduce statistical power
2. Bias in estimation of parameters (SE)
3. Reduce representativeness of the sample