Prob/Stats 4.1.4 Observational Studies and Experiments 4-39

Prob/Stats 4.1.4 Observational Studies and Experiments 4-39

Consider these actual newspaper headlines.

You will try to determine if they are an observational study or an experiment. If an experiment is possible, you will try to determine a design and include a placebo, if possible. If surveys were necessary, you will list a potential source of bias in the question(s), and a potential difficulty in getting a representative sample from the population.

Multiple Choice

STUDY STICKS IT TO TRADITIONAL BACK CARE.

Acupuncture—real and fake—gets better results for pain than the usual treatments.

Do you think this headline came from an observational study or an experiment?

• Observational Study

• Experiment

Open Ended

STUDY STICKS IT TO TRADITIONAL BACK CARE.

Acupuncture—real and fake—gets better results for pain than the usual treatments.

Because a “fake” acupuncture placebo is mentioned this most certainly was an experiment.

Outline a possible experimental design including a placebo if possible. Again, don't worry about being wrong! This is just to get you thinking...

Here is an example of what an experiment could look like (compare to what you said on the last slide and note that there are multiple correct answers!) [Click "next" to see each bullet]

• A group of volunteers with back pain were randomly distributed between at least 3 test groups.

• One group received traditional medical care, another got acupuncture treatments, and the last got fake acupuncture (the placebo).

Open Ended

Could we use a study with the experiment to measure back pain? What bias could occur?

The difference in pain reported by each patient was likely averaged for each group and the averages compared using a study.

• Measuring the subjects’ pain was likely done by survey, i.e., “Rate your back pain today from 1 to 10…”

• The physical surroundings of the interview or the compassion level of the interviewer may be a source of bias.

Multiple Choice

Here's another real headline:

MARITAL STRIFE A HEART WRECKER? Bad marriage can increase risk of coronary disease, researchers say.

Do you think this headline came from an observational study or an experiment?

• Observational Study

• Experiment

Open Ended

MARITAL STRIFE A HEART WRECKER? Bad marriage can increase risk of coronary disease, researchers say.

This is an observational study. A large group of people were given questionnaires including questions about their marriage and their health. This cannot be an experiment because a randomly selected group of people would have to be forced into “bad” marriages and their heart health compared to a group forced into “not bad” marriages. Sounds like a reality TV show though, doesn't it?

What are some confounding variables in this study? (In other words, the headline says Bad marriage = coronary disease, but not everyone who has coronary disease is in a bad marriage and vice-versa. So what are some other explanations for this relationship?)

Multiple Choice

Some confounding variables from the last slide could be: Perhaps people under a great deal of stress tend to have bad marriages and tend to have heart problems. Maybe people with addiction problems tend to have bad marriages and tend to have heart problems. Again, more than one correct answer is okay here.

What is the population for this study? Here is the headline again: MARITAL STRIFE A HEART WRECKER? Bad marriage can increase risk of coronary disease, researchers say.

• People with coronary disease

• People in bad marriages

• Married people in the US

• Married Couples in the world

Multiple Choice

Were you surprised by the last slide's answer? It just says married people, so it is literally making a claim about everyone in the world that is married. Do you think it would be easy to get a random sample of the world's married population?

• Yes, so easy!

• No, it would be extremely difficult.

Trying to get a random sample of responses from the world’s married population would be nearly impossible. It is likely the respondents had something in common like an employer or health plan. It is not very representative of all married couples, that is for sure!

• Maybe having a bad marriage or heart condition made people more likely to respond to the survey?

• Also, what constitutes a “bad” marriage? That is very subjective! Be wary when you see headlines like these!

Multiple Choice

Here's another real headline:

BREASTFEEDING MAY CUT BREAST CANCER RISK.

Women with a family history of breast cancer who have ever breastfed reduce their risk of getting premenopausal breast cancer by nearly 60%, according to a new study.

Do you think this headline came from an observational study or an experiment?

• Observational Study

• Experiment

Open Ended

BREASTFEEDING MAY CUT BREAST CANCER RISK.

Women with a family history of breast cancer who have ever breastfed reduce their risk of getting premenopausal breast cancer by nearly 60%, according to a new study.

This is an observational study. A large group of women were given questionnaires including questions about breastfeeding and whether they have had breast cancer. This cannot be an experiment because a randomly selected group of women would have to be forced to breastfeed their children and their breast cancer rates compared to another group forced to not breastfeed their children. Yikes!

What are some confounding variables in this study? What are other reasons why women who breastfeed might have a reduced Cancer risk?

Multiple Choice

Some confounding variables from the last slide could be: Perhaps women who breastfeed are less likely to drink alcohol or smoke. Maybe women who breast feed eat healthier foods than those who do not.

What is the population for this study? Here is the headline again: BREASTFEEDING MAY CUT BREAST CANCER RISK.

Women with a family history of breast cancer who have ever breastfed reduce their risk of getting premenopausal breast cancer by nearly 60%, according to a new study.

• People with breast cancer

• Women with a family history of breast cancer

• Women who breastfeed

• Women with children

Multiple Choice

Again, this headline is very vague in that it just says "women with a family history of breast cancer..." and doesn't specify where, so it could be any women in the world with a family history of breast cancer. Do you think it would be easy to get a random sample of these women?

• Yes, so easy!

• No, it would be extremely difficult.

Again, trying to get a random sample of responses from the world’s population of women with a family history of breast cancer would be nearly impossible.

• It is likely the respondents had something in common like an employer or health plan.

• Maybe the type of people who take (or have) the time to breastfeed their children, also takes the time to fill out health surveys, so that group was over represented in the sample.

Open Ended

The final question is: What do you think the point of today's lesson was?