Lesson test: Where the data comes from

Sign in

Libre University uses your GitHub account. Signing in is only needed to sit a final test, so the score is kept on your profile.

Where the data comes from

1.[2p]
The 1936 Literary Digest poll collected 2.4 million responses and missed by 17.8 points. What does that show?

Coverage and non-response bias do not shrink with sample size

Correct

The sample was too small to detect the true margin

The answer is: Coverage and non-response bias do not shrink with sample size

The standard error formula does not apply to proportions

The answer is: Coverage and non-response bias do not shrink with sample size

Roosevelt's support changed between the survey and the election

The answer is: Coverage and non-response bias do not shrink with sample size

2.[3p]
Match each term to what it names.

Target population
Sampling frame
Coverage error
Non-response bias

everyone the question is about
the list actually sampled from
responding is related to the answer
the gap between frame and population

Show the answer

Target population: everyone the question is about Sampling frame: the list actually sampled from Coverage error: the gap between frame and population Non-response bias: responding is related to the answer

3.[3p]
A survey gets a 25 per cent response rate. Support is 64 per cent among respondents and 40 per cent among non-respondents. What is the true population percentage?

CorrectNot quite: 46

4.[2p]
A cluster sample uses clusters of 20 with intraclass correlation $ρ = 0.05$ . What is the design effect $1 + (m - 1) ρ$ ?

CorrectNot quite: 1.95

5.[3p]
Strata hold fractions 0.8 and 0.2 of a population with standard deviations 5 and 10. Under proportional allocation with $n = 100$ , the variance is $\frac{1}{n} \sum W_{h} σ_{h}^{2}$ . What is it?

CorrectNot quite: 0.4

6.[2p]
A stratified sample with proportional allocation removes the between-stratum part of the variance from the estimate.

True

Correct

False

The answer is: True

7.[3p]
Which of these are reasons a very large found dataset can be worse than a small random sample?

Select all that apply

Selection can be correlated with the quantity measured, and that term does not shrink with $n$

Correct

Its size makes readers trust it more than its design warrants

Correct

The reported standard error describes only the sampling error

Correct

Large datasets always have larger measurement error per record

The answer is: Selection can be correlated with the quantity measured, and that term does not shrink with $n$, Its size makes readers trust it more than its design warrants, The reported standard error describes only the sampling error

8.[2p]
What does simple random sampling actually guarantee?

Every unit has a known chance of inclusion, which is what licenses the later formulas

Correct

The sample resembles the population on every variable

The answer is: Every unit has a known chance of inclusion, which is what licenses the later formulas

No two samples give the same estimate

The answer is: Every unit has a known chance of inclusion, which is what licenses the later formulas

The estimate equals the population value

The answer is: Every unit has a known chance of inclusion, which is what licenses the later formulas

9.[2p]
Put these steps of a national survey in the order they happen.

Weight the responses back to known population totals
Obtain or build a sampling frame
Collect responses and record the response rate
Draw the sample by a random mechanism
Define the target population

Show the answer

a, b, c, d, e

1.[2p] The 1936 Literary Digest poll collected 2.4 million responses and missed by 17.8 points. What does that show?

2.[3p] Match each term to what it names.

3.[3p] A survey gets a 25 per cent response rate. Support is 64 per cent among respondents and 40 per cent among non-respondents. What is the true population percentage?

4.[2p] A cluster sample uses clusters of 20 with intraclass correlation ρ=0.05. What is the design effect 1+(m-1)ρ?

5.[3p] Strata hold fractions 0.8 and 0.2 of a population with standard deviations 5 and 10. Under proportional allocation with n=100, the variance is 1n∑Whσh2. What is it?

6.[2p] A stratified sample with proportional allocation removes the between-stratum part of the variance from the estimate.

7.[3p] Which of these are reasons a very large found dataset can be worse than a small random sample?

8.[2p] What does simple random sampling actually guarantee?

9.[2p] Put these steps of a national survey in the order they happen.

1.[2p]
The 1936 Literary Digest poll collected 2.4 million responses and missed by 17.8 points. What does that show?

2.[3p]
Match each term to what it names.

3.[3p]
A survey gets a 25 per cent response rate. Support is 64 per cent among respondents and 40 per cent among non-respondents. What is the true population percentage?

4.[2p]
A cluster sample uses clusters of 20 with intraclass correlation $ρ = 0.05$ . What is the design effect $1 + (m - 1) ρ$ ?

5.[3p]
Strata hold fractions 0.8 and 0.2 of a population with standard deviations 5 and 10. Under proportional allocation with $n = 100$ , the variance is $\frac{1}{n} \sum W_{h} σ_{h}^{2}$ . What is it?

6.[2p]
A stratified sample with proportional allocation removes the between-stratum part of the variance from the estimate.

7.[3p]
Which of these are reasons a very large found dataset can be worse than a small random sample?

8.[2p]
What does simple random sampling actually guarantee?

9.[2p]
Put these steps of a national survey in the order they happen.