quiz 3/13

Question 1

What are the three main types of distance measures covered?

Answer 1

1. Distances for numerical data
2. Distances for proportions
3. Distances for presence-absence data

Question 2

When should you use distance measures for presence-absence data?

Answer 2

• When data is organized as a set of features for different individuals/samples/populations.
• Examples include:
  
  • Whether species are present (1) or absent (0) in different locations.
  • Whether viruses express certain genetic markers.
  • Whether bodies of text contain specific words.

Question 3

How is presence-absence data summarized?

Answer 3

• In a contingency table with counts:
  
  • a = Features present in both samples
  • b = Features present in Sample A but absent in Sample B
  • c = Features present in Sample B but absent in Sample A
  • d = Features absent in both samples
  • n = Total features (a + b + c + d)

Question 4

How can similarity be derived from a distance measure?

Answer 4

• Similarity is often computed as:Similarity = 1 - Distance
• This works for presence-absence data and proportions.

Question 5

What does the Mantel Randomization Test do?

Answer 5

• It compares two distance matrices to see if their distances are correlated.
• Tests if one type of distance (e.g., economic indicators) relates to another (e.g., health outcomes).

Question 6

What are some real-world questions the Mantel Test can answer?

Answer 6

• Do economic indicators impact health outcomes?
• Does geographic distance affect genetic similarity?
• Do news announcements influence stock market fluctuations?

Question 7

How does the Mantel Test determine significance?

Answer 7

• Compute the similarity index for all possible random permutations of one of the matrices.
• Compare the actual similarity to the distribution of randomized similarities.
• If the observed similarity is much higher than expected by chance, we reject the null hypothesis.

Question 8

What do the Mantel Test results tell us?

Answer 8

• If two distance matrices are significantly correlated, their distances are related.
• If not, the patterns in one matrix do not predict patterns in the other.

Question 9

What is the formula for the Simple Matching Coefficient (SMC)?

Answer 9

a+d / a+b+c+d

Counts both presence (a) and absence (d) as agreement.

Question 10

What is the formula for Jaccard Similarity?

Answer 10

a/(a+b+c)

Does not count shared absences (d) as similarity.

Question 11

What is the formula for Sørensen-Dice Similarity?

Answer 11

2a/(2a + b +c)

Gives twice the weight to shared presences (a) compared to Jaccard.

Question 12

What is the formula for Ochiai Similarity?

Answer 12

a / (sqrt((a+b) * (a+c)))

Similar to Jaccard but accounts for sample sizes using a square root.

Question 13

When should i use ochiai similarity eqn?

Answer 13

When massive difference in samples and groups

Question 14

when do you want to use sorensen dice?

Answer 14

when a is small

Question 15

when do you want to use jarrcard?

Answer 15

when a is not small

Question 16

when do you want to use simple matching?

Answer 16

when you think the absences (d) mean something