Quiz · Hard
Fifteen calculation and interpretation questions on sample size, confidence intervals, z-scores, market sizing, NPS and hypothesis testing.
A pollster wants a 95% confidence level, p = 0.5 and a margin of error of 5%. What minimum sample size is needed?
196
385
769
1,537
n = 1.96^2 x 0.25 / 0.05^2 = 384.16, rounded up to 385.
Halving the margin of error from 4% to 2% requires the sample size to:
Double
Quadruple
Triple
Stay the same
n is proportional to 1/E^2, so halving E multiplies n by 4.
A survey of 625 people finds 40% prefer Brand X. What is the 95% margin of error?
About 2%
About 1.96%
About 7.8%
About 3.9%
E = 1.96 x sqrt(0.4 x 0.6 / 625) = 1.96 x 0.0196 = 0.0384, or about 3.8-3.9 points.
In that survey, which interval is the 95% confidence interval for the share preferring Brand X (rounded)?
36.2% to 43.8%
20% to 60%
38% to 42%
30% to 50%
40% plus or minus 3.8 points gives 36.2% to 43.8%.
A product sells to 500,000 households per year at an average of $120 each in the serviceable market. A firm expects to win 8% of it. What is its SOM revenue?
$60 million
$48 million
$4.8 million
$4.8 billion
SAM = 500,000 x 120 = $60 million. SOM = 8% x $60 million = $4.8 million.
A company has 18% market share. Its largest rival has 30%. What is the company's relative market share?
1.2
1.67
0.18
0.6
Relative share = 18/30 = 0.6. Below 1 means it is not the market leader.
In a normal distribution of weekly spend (mean $80, SD $15), roughly what share of customers spend between $50 and $110?
68%
95%
99.7%
50%
$50 and $110 are 2 SDs below and above the mean, so about 95%.
A customer's spend is $95 where mean is $80 and SD is $15. What is the z-score?
0.5
1.5
1
15
z = (95 - 80)/15 = 1.
Interviewers must complete 400 interviews: 55% women and 45% men to match the population, choosing respondents on the street. This is:
Quota sampling
Stratified random sampling
Cluster sampling
Simple random sampling
Fixed group quotas filled by non-random selection define quota sampling.
A sample of 1,000 is drawn from a customer list that excludes anyone who joined in the last year. What problem does this create?
Sampling error only
Social desirability bias
Undercoverage in the sampling frame
Acquiescence bias
The frame does not cover the whole population, so the sample cannot represent new customers.
An A/B test p-value is 0.20 with alpha = 0.05. What should the researcher conclude?
Fail to reject the null: not enough evidence of a difference
Reject the null: significant
Accept the variant
The test has Type I error
p > alpha, so the observed difference could easily be chance. It does not prove there is no difference.
A company concludes that a new package works when it does not (false positive). This is:
Measurement error
Type II error
Sampling error
Type I error
Type I: rejecting a true null hypothesis. Type II: missing a real effect.
Survey of 200 customers: 120 are satisfied. What is the sample proportion and its standard error?
0.60 and 0.35
0.60 and 0.035
0.40 and 0.035
0.60 and 0.0035
p = 120/200 = 0.60; SE = sqrt(0.6 x 0.4 / 200) = sqrt(0.0012) = 0.0346, about 0.035.
Which statement about regression coefficients is correct?
They prove cause and effect
They estimate the change in the dependent variable for a one-unit change in an independent variable, holding others constant
They are always positive
They require a census
Regression quantifies relationships but cannot alone prove causation.
A firm surveys 100 customers (30 detractors, 30 passives, 40 promoters). What is the NPS?
+30
+40
-10
+10
NPS = 40% promoters - 30% detractors = +10.