32. Statistical test selection, Interpretation of the evidence

Question 1

What are the criteria for selecting a statistical test for analyzing data?

Answer 1

• Scale of measurement (nominal, ordinal, interval, ratio)
• Number of groups (two or more)
• Independent or related samples
• Assumptions of the statistical test

These criteria help determine the appropriate statistical test for a specific study.

Question 2

Define effect size in the context of research interpretation.

Answer 2

The size of the change or degree of association attributed to a health intervention

Effect size is crucial for understanding the practical significance of research findings beyond mere statistical significance.

Question 3

Explain the relationship between effect size, statistical significance, and clinical significance.

Answer 3

• Statistical significance: results due to chance or real trends
• Effect size: magnitude of the effect
• Clinical significance: practical importance of the effect

Statistical significance does not always imply clinical significance.

Question 4

What are the key determinants of statistical power?

Answer 4

• Sample size
• Effect size
• Significance level

Statistical power is the probability of correctly rejecting the null hypothesis.

Question 5

True or false: A power of 0.95 indicates a high likelihood of detecting an effect.

Answer 5

TRUE

A power of 0.95 means the analysis will correctly detect the effect 95 times out of 100.

Question 6

When should a t test be used instead of a z test?

Answer 6

When n < 30

This is due to the t test being more appropriate for smaller sample sizes.

Question 7

What type of statistical tests are used for nominal and ordinal data?

Answer 7

• Non-parametric tests
• Distribution-free tests

These tests require few or no assumptions about population distributions.

Question 8

What type of statistical tests are used for interval or ratio data?

Answer 8

• Parametric tests (e.g., z, t, F)

These tests require assumptions such as normality and equal variance.

Question 9

If measuring outcomes on a 5-point ordinal scale with two groups, which test should be selected?

Answer 9

Mann-Whitney U test

This test is appropriate for comparing two independent groups with ordinal data.

Question 10

If there are three groups and the outcome is ordinal, which test should be selected?

Answer 10

Kruskal Wallis H test

This test is used for comparing three or more independent groups with ordinal data.

Question 11

What is the significance level typically used to determine statistical significance?

Answer 11

p < 0.05

This indicates a 5% probability that the results are due to chance.

Question 12

What is the power of a statistical analysis defined as?

Answer 12

Power = 1 - β (probability of a miss)

This formula helps in understanding the likelihood of detecting a true effect.

Question 13

What is the best defense against low statistical power?

Answer 13

A good-sized sample

Larger sample sizes improve the likelihood of detecting true effects.

Question 14

In clinical decision-making, what is generally more acceptable: Type I error or Type II error?

Answer 14

Type I error (false alarm)

Clinicians tend to minimize misses (Type II errors) even if it leads to unnecessary interventions.

Question 15

What is the relationship between effect size, sample size, and decision making?

Answer 15

Clinician decision procedures are similar to hypothesis-testing procedures

Both involve making decisions based on uncertain information.