- On April 4, 2017, Gallup published poll results on its web site under the headline, "Affordable Care Act
Gains Majority Approval for First Time."
Of 1,023 adults surveyed, 55% of them responded "approve" to
the question, "Do you generally approve or disapprove of the 2010 Affordable Care Act, signed into law
by President Obama, that restructured the U.S. healthcare system?" The article also notes that the ACA
had never before showed majority support in Gallup polling, but that 48% of the sample said "approved"
the first time the current version of the question was asked in November 2012.
If this was a simple random sample of the 1023 adults in 2017,
what would the conservative 95% confidence interval
margin of error be?
- approximately 5%
- approximately .03%
- approximately 3.13%
- Gallup gives a 95% confident margin of error of plus or minus 3%
for the 2012 poll, which had 48% of the sample "approved." So the lower and upper boundaries
for the confidence interval is
48% -3% = 45% to 48 % + 3% = 51%
the lower and upper boundaries for 95% confidence intervals for the "approve" results for the 2017
poll, which had 55% of the sample "approved" and a margin of error plus or minus 4%
- Assume adults 18 and over in the U.S. are the population of interest for Gallup's poll. If we account for
random error in the sample using the stated margin of error, is the headline "Affordable Care Act Gains
Majority Approval for First Time" accurate with regard to the population?
(Note: Here majority means more than 50%.)
Use the confidence intervals to answer two questions:
First, was it likely a majority in 2017?
Second, could it have been a majority earlier---in 2012?
- For a simple random sample at the 95% confidence level, what sample size would be required to achieve a plus or minus 1% margin of error?
- Gallup specifically targets both landline and cellphone users in its polls. Are there any voices that are left out?
- What was the main point of Fisher's experiment on the Lady Tasting Tea?
- sample size and random nature of a representative selection is what is important---not the
percentage of the overall population (like for chicken soup)
- we can't assume that unusual data is incorrect (like the negative
growth rate for US population in 1918)
- Leonardo DaVinci asserted that we are squares (in terms of armspan and
- SAT Scores are loosely correlated to college GPA, but don't cause them
- statistical significance can be obtained by deciding in advance what
level of confidence we will accept as persuasive
and we can collect data in such a way that we can make reasoned inferences
Regarding HIV testing,
- the large numbers of HIV negative people can have
false positives and make what seems like an accurate test percentage-wise problematic
- A positive result becomes relatively meaningless because one only has a small chance of actually having HIV.
However, what it does help reveal is to change the probability that a person is HIV-positive
from roughly 3 in 1000 to roughly 1 in 4.
- Testing the entire US population is what leads to the (unintended) problems. Testing other populations would require
a different analysis
- all of the above
- Which did you find most compelling about the "price of life" readings
- unintended consequences of HIV testing the entire US population
- unintended consequences of raising plane tickets to improve air traffic safety & car accident statistics
- cost of pap smears
- asbestos removal
- lack of education and poverty can lead to poorer options/decisions regarding personal health (and correlation to an earlier death)
What percentage of us population has a phone?
How many adults adults do not use the internet. Demographics of Internet and Home Broadband Usage in the United States.