Statistics

Question 1

What types of data are there?

Answer 1

Quantitative and Qualitative

Quantitative (categorical)
1. Nominal- unordered, but mutually exclusive groups like gender, blood group

2. Ordinal- ordered and mutually exclusive like ASA grades, Mallampati

Qualitative (numerical):

• discrete limited numbers (days of annual leave)
• continuous - endless possible numbers (height, weight)

Question 2

What is a Type 1 Error?

Answer 2

Rejecting the null hypothesis when there is there is no statistical difference.

False positive.

Question 3

What’s a type 2 error?

Answer 3

Accepting the null hypothesis even though there is a statistical difference. Null hypothesis is false.

False negative.

Question 4

Causes of Type 1 errors?

Answer 4

Bias, Confounding, multiple hypothesis testing.

Question 5

What’s the chance of making a type 1 error?

Answer 5

The same as p value

Question 6

What causes type 2 errors?

Answer 6

Small sample sizes, excessive variance.

A degree of power is calculated - 0.8 is used. This means 80% probability study will yield a significant result.

Question 7

What test can be used to compare categorical, nominal data Ie. gender:

1. One sample comparison
2. Two group comparison
3. Multiple group comparison

Answer 7

1. One sample - Chi-squared test
2. Two groups/ multiple groups:
   - unpaired (chi-sq with Yates correction)
   - paired (McNemars)

Question 8

What tests can be used to compare non-parametric, ordinal ie, ASA groups:

1. One sample
2. Two groups
3. Multiple groups

Answer 8

1. One sample - Wilcox signed rank
2. Two groups:
   - paired - Mann Whitney U Test
   - unpaired - Wilcox matched
3. Multiple groups:
   - paired - Friedman
   - unpaired - Kruskal-Wallis

Question 9

Statistical tests for parametric, continuous data Ie. weight:

1. One sample
2. Two groups
3. Multiple groups

Answer 9

1. One sample t-test
2. Students T-test (paired and unpaired)
3. ANOVA (paired and unpaired)

Question 10

What type of data is parametric vs non-parametric?

Answer 10

Parametric is continuous quantitative data. Centred about the mean and normal distributed.

Non-parametric is qualitative data both nominal and ordinal. Centred about the median and is can be any type of distribution.

Question 11

What is recall bias?

Answer 11

Affects retrospective studies when participants selectively recall details of the past.

Question 12

What is response bias?

Answer 12

Participant answers questions in a way they thing the researcher is expecting, not actual belief/ experience.

Question 13

Ascertainment bias?

Answer 13

Researcher is not blinded to the participants intervention or exposure.

The knowledge influences how participant is treated/ included or measured.

Question 14

Hawthorne Effect?

Answer 14

When a participant is aware they are being observed and change behaviour consciously.

Ie changing diet because they have to keep a food diary.

Question 15

Attrition bias?

Answer 15

In a study with two randomised treatment arms,

If participants drop more out of one arm of the study,

May bias data by not adequately representing the original study population.

Question 16

What’s the difference between simple and stratified randomisation?

Answer 16

Simple randomisation allocates patients randomly with toss of the coin or random number tables.

Stratified randomisation takes into account that confounding factors may not be equally distributed this way. Takes a subgroup of patients with confounding factors and and randomised within these groups.