Sampling Variation, Bias, Confounding

Question 1

What is the sampling distribution?

Answer 1

The distribution of a statistic across an infinite number of samples.

Question 2

What is the 95% confidence interval?

Answer 2

• Range that is likely to contain the mean of the population values 95% of the time.
• I.e. Range in which you are 95% sure that the mean of the population values lie.

Question 3

What is the 95% Confidence Interval based on?

Answer 3

Gaussian distribution (of the means of the samples)

Question 4

The 95% confidence interval is wider the…

Answer 4

… greater the variation in population values.

… smaller the size of the sample used to calculate it.

Question 5

What is a confidence interval?

Answer 5

Estimate of the precision of the observed values in the sample.

Question 6

What are the 2 main categories of bias?

Answer 6

1. Selection bias

2. Information bias

Question 7

What are the reasons for selection bias and the impact of these?

Answer 7

1. Errors in generalisability (external validity)
   
   • Study Ps are drawn from a sampling frame that is not representative of the general population .
2. Errors in comparability (internal validity).
   
   • The groups being compared are not from the same population.
   • One of the groups being studies is not representative of the sampling frame from which it was drawn (e.g. Low response rate to a survey, high attrition rate).

Question 8

What are the different types of information bias?

Answer 8

1. Differential recall error (especially in case-control studies)
   
   • e.g. Parents of ill child better recall than parents of unaffected child.
2. Differential observer of interviewer error
3. Differential measurement error
4. Differential mis-classification

Question 9

What is a confounding variable?

Answer 9

An extraneous variable that correlates with both the dependent and independent variable.

Question 10

What is the biggest confounding variable and what does this arise from?

Answer 10

• Age

- Demographic differences (i.e. Rapidly expanding, expanding, stationary and contraction).

Question 11

How can confounding variables be controlled for (e.g. Standardising for effect of age in death rate of 2 pops with different demographics)?

Answer 11

• Adjust with direct standardisation.
  
  • adjust number of people in each age category so that they are = in 2 samples.
• Adjust with indirect standardisation.
  
  • adjust death rate (of each age group) so that it is = in 2 samples.