Seminar #21Karen Jakubowski

Finding 95% Confidence Intervals: An approximate 95% confidence interval for

unknown population proportion p is based on sample proportion p-hat from a random sample of size n

= sample proportion +/- 2 standard deviations

Article #1: The White Coat Syndrome

Some people exhibit a psychophysiological response to seeing doctors or other medical professionals

This has been termed the “White Coat Syndrome” or “White Coat Hypertension” Patients suffer from hypertension only when in the

presence of a person in a “white coat.”

Study design: Study involved 419 patients who exhibited

hypertension while at their doctor’s office. They were provided with portable blood pressure-

measuring devices that measured their levels outside of the office.

Results: 26% (109) of the patients suffered from

hypertension only while visiting their physician.

What are some possible reasons that individuals exhibit the “White Coat Syndrome?” Fear of bad news Knowledge of experiences of friends/family Embarrassment (ie: have not been taking

medication regularly or following past directions)

Finding 95% Confidence Interval .26 +/- 2SqRoot (.26(1-.26)) / (419)

= (.217, .303)

Article #2: Nine Percent of U.S. Children Age 8 to 15 Meet Criteria For Having ADHD 8.7% of U.S. children age 8 to 15 meet

diagnostic criteria for ADHD fewer than half receive treatment

ADHD is characterized by hyperactivity, impulsive behavior, and an inability to pay attention to tasks.

Study Design: Study involved 3,082 children

sample population designed to represent the entire population of 8 to 15-year-olds in the U.S.

Parents/caregivers provided information about their child’s ADHD symptoms and medical history, as well as sociodemographic details via phone interview.

Study Design: How could this study design have been

flawed? Some parents may not have been entirely

truthful about their child’s ADHD symptoms or medical history.

Bias from parents’ inaccurate memories about when their child first displayed symptoms of ADHD or the severity of the symptoms.

Finding 95% Confidence Interval 8.7% (268) of the 3,082 children studied

fulfilled criteria for ADHD. .087 +/- 2SqRt ((.087(1-.087)) / 3082

= (.077, .097)

Finding 95% Confidence Interval 47.9% of the children who met ADHD

criteria (268 children) had been previously diagnosed with the condition.

Before we calculate the interval, do you think that the children meeting ADHD criteria who had been diagnosed were in a minority?

Finding 95% Confidence Interval .479 +/- 2SqRt ( (.479 (1-.479)) / 268 )

= (0.42, 0.54)

Since the values in our interval surround .5 , we cannot be certain that the children who were diagnosed with ADHD were in a minority.

Article #3: Surgeons With Video Game Skill Appear To Perform Better In Simulated Surgery Skills Course Study involved 33 surgeons: 12 attending

physicians and 21 residents Asked about their video game-playing habits, then

assessed on their performance at the Rosser Top Gun Laparoscopic Skills and Suturing Program A 1.5 day course that scores surgeons on time and errors

during simulated surgery skills.

Results: Surgeons who had played video games in the past

for more than 3 hours/week made 37% fewer errors, were 27% faster, and scored 42% better overall than surgeons who never played video games.

Current video game players made 32% fewer errors, were 24% faster and scored 26% better overall than non-players.

Surgeons in the top 1/3 of gaming skill made 47% fewer errors, performed 39% faster, and scored 41% better overall than those in the bottom 1/3.

Does anyone notice a problem …?

From the information provided, we cannot find 95% confidence intervals! The data was summarized in quantitative terms, but we

were not given any mean values, just percentages comparing how much higher the mean for one group is compared to another.

Therefore we cannot set up a confidence interval around a proportion in this example.

Article #4: U.S. College Students’ Exposure to Tobacco Promotions: Prevalence and Association With Tobacco Use This study assessed college students’

exposure to the tobacco industry marketing strategy of sponsoring social events at bars, nightclubs, and college campuses.

Study design: Data came from the 2001 Harvard College

Alcohol Study - a random sample of 10,904 students enrolled in

119 “nationally representative” 4-year colleges and universities.

Study design: Questionnaires were mailed to 21,055 students in

February 2001. 3 mailings were sent within 3 weeks: the

questionnaire, a reminder, and a second questionnaire.

Responses were anonymous, and cash prizes were awarded to encourage responses.

What types of questions should the questionnaire have included?

Study design: The questionnaire assessed students’:

demographics (ie: age, sex, race, GPA) tobacco, alcohol, and marijuana use

Tobacco use was defined as having (in the past 30 days): smoked a cigarette, cigar, pipe, or bidi (a small

hand-rolled often flavored cigarette made in India)

used smokeless tobacco

Study design: What aspects of this study reduced bias?

Using a large sample Using a sample that was representative of the

larger college student population Anonymous questionnaire

Study design: What aspects of this study could have

caused bias? Non-response bias Not answering the questions truthfully Wording of the questions

Results: 52% (5,670 students) responded to the

questionnaire

Do you think that 52% a good response rate? Does it provide enough information to allow

accurate inferences to be made about the larger population of college students?

Results: The effect of exposure of tobacco promotions

differed by the age at which students first began smoking

Out of the 78% (8482) of students who did not smoke regularly before 19 years of age (approximately the age most students enter college) the current smoking rate was 23.7% for students who had attended a promotional event 11.8% for students who had not attended an event

Finding 95% Confidence Interval For the 23.7% of students who had not smoked before

age 19, but were current smokers and had attended a promotional event:.237 +/- 2SqRt ( (.237(1-.237)) / 8482)

= (.227, .246)

For the 11.8% of students who had not smoked before age 19, but were current smokers and had never attended a promotional event:.118 +/- 2SqRt (.118(1-.118)) / 8482

= (.111, .125)

What does it mean? Since the confidence intervals for the two

separate groups do not overlap, the data suggests that one population proportion is higher than the other.

Results: For the 22% (2334) of students who

smoked regularly before 19 years of age, there was no significant difference between the percentage of students who had or had not attended a tobacco promotional event. 77.5% vs 72.2%, respectively

Conclusion: Tobacco promotional events may

encourage previously non-smoking college students to begin smoking, or current smokers to continue smoking.

The End!