Module 2

Question 1

Define null hypothesis (H0)

Answer 1

Hypothesis of no effect/nothing happening

Question 2

Define alternative hypothesis (HA)

Answer 2

Describes effect expected to see

Question 3

Hypothesis Testing Steps

Answer 3

1. State H0 and HA
2. Compute test statistics, compare data to H0 (how well sample results fir if H0 is true)
3. Determine P-value
4. Draw appropriate conclusions

Question 4

Define null distribution

Answer 4

Probability distribution of a test statistic value when a random sample is taken from a hypothetical population for which the null hypothesis is true

Question 5

Type 1 error (alpha)

Answer 5

Rejecting a true null hypothesis (false positive)
- 5% chance of type 1 error usually

Question 6

Type 2 error (beta)

Answer 6

Failing to reject a false null hypothesis (false negative)
- Do not usually know this value
- The smaller B, the greater the power of the test (power = 1 - B)

Question 7

Power depends on:

Answer 7

• How different the truth is from the null hypothesis
• Type 1 error rate
• Precision (sample size)

Question 8

If the test statistic we obtain from our sample leads to:

Answer 8

• p-value < alpha = rejection of H0, there is a significant difference
• p-value >= alpha = fail to reject H0

Question 9

Decreasing type 1 error/alpha:

Answer 9

makes it harder to reject H0
- decreases type 1 error rate
- increases type 2 error rate
- decreases power

Question 10

What are the ways to decrease sampling error?

Answer 10

• Replication
• Balance
• Blocking
• Extreme treatments

Question 11

Replication

Answer 11

• Application of treatment to multiple, independent experimental subjects or units
• To measure the precision of the estimate
• Avoid pseudoreplication by correctly identifying independent replicates

Question 12

Balance

Answer 12

Equal # of units in each treatment
- Minimizes the SE associated with the treatments. Larger samples are always better, but a balanced design allocates sampling effort optimally

Question 13

Blocking

Answer 13

• Helps to account for extraneous variation by dividing experimental units into groups
• Both treatments in both situations
• Cannot limit variation, so must account for it

Question 14

If you know what confounding variables might be, use:

Answer 14

blocking

Question 15

If you don’t know what the confounding variables might be, use:

Answer 15

random assignment

Question 16

Extreme treatments

Answer 16

Bigger manipulation usually leads to a bigger response
- Responses are not always linear