Session 5:

​

Data and

Statistics

ICEBREAKER & CHECK IN

Why are data and statistics important for public health research?

What is Data?

• Data are bits of information about

individuals

Examples:

1. patient demographics - occupation,

ethnicity, area of residence, income

• patient outcomes in clinical trials - response to treatment, adverse events, assay results
• genetic sequencing - Ancestry and 23andMe, also a huge area of focus for bioinformatics
• public health data - COVID-19 case numbers, positivity rates, hospitalizations, variant frequencies

What is Data Science?

• Data is often NOISY and there’s a LOT of it
• Data science - area of study that attempts to make meaning out of this “big data”
• Why is it important?
• The digitization of the modern

world (i.e. the internet) has made

large amount of data available to

individuals

• Collecting data is crucial for

answering questions and solving

problems - data-driven science

What is Statistics?

• One of the most popular data science methods
• We usually only have data on a subset of individuals out of an entire population, or a sample
• Statistics summarize this sample data in a

meaningful way

• Enables us to make inferences without

needing data from the entire population

• Allows for comparisons between groups
• Allow us to make sense of experiments

What is Statistics?

• You’re probably familiar with some of these
• Mean
• Median
• Standard Deviation
• IQR
• They give us information about a certain group/sample
• There are also statistical methods for COMPARING groups
• Confidence Intervals
• Hypothesis Testing

Statistics Terminology

What is a “p-value?” What about a “confidence interval?”

P-value

• The probability that there is some non-random explanation for the things we observe in a data set
• In experiments, used to support the thought that our experiment actually caused certain effects
• In other words: what are the odds that we would observe this result from chance alone?

P-values, cont’d.

• The general baseline that you’re aiming for is a p-value of <0.05
• That means that there’s a <5% chance of this thing happening completely at random
• You want a lower p-value, because it means that there’s a lower chance that this thing happened randomly = higher chance that your “treatment” is what caused the change

How do you get a p-value?

• Statistical tests known as t-tests or z-tests, which we won’t cover at the moment
• Calculated by comparing an experimental group and a control group (null)

Confidence Interval

​

• A range of values that we think capture the “true” value for a statistic (e.g. the mean for an entire population) some percentage of the time if we sampled MANY times
• Usually aim for a 95% confidence interval
• Also has a formula, but we won’t go into it

Confidence Interval

​

• If they don’t overlap, we know there is a statistically significant difference between the groups being compared

Confidence Intervals

The picture shows early results on Pfizer’s vaccine before it was sponsored by the company

Confidence Intervals

This picture shows a comparison between different vaccines, again before they were sponsored by companies.

Keep in mind the confidence intervals are represented by shading here instead of error bars

What is Data Visualization?

• Ok - we have all this data summarized with statistics
• What do we do with it?
• Data visualization is displaying

those pieces of information so

we can see their insights

• KEY to scientific

communication

• Handy tools:
• Excel - small amounts of data
• Programming languages like R and Python - big data

Data Massaging

(aka “Fudging the Numbers”)

• Sometimes, people may manipulate/adjust data in order to show the results they want in their studies or papers
• What might be a few reasons people would do this?
• Funding - some institutions provide resources to their labs based on successful results or paper publications
• Agenda - some people might want to just push an agenda and thus cheat their results accordingly
• etc.

Case Study: Theranos

• Theranos: company that supposedly offered highly informative blood testing that would revolutionize the medical industry
• The company had at least a billion dollar evaluation $$$
• When investigations were performed: administration pressured employees to fake data and keep stuff secret
• Became a WIDELY publicized scandal

Takeaways

• ALWAYS take reported statistics and data with a grain of salt
• Make sure that the data has been mined responsibly and that a random sample is being used if possible
• Some individuals may cherry pick their data to push an agenda or fake promising results
• Most peer-reviewed sources will be safe from this stuff, so rest assured those will be reliable

Sources Activity

Mentees: Take a look at your collected sources and analyze the data they provide (if any). Discuss how you can interpret these findings.

​

Community

Interviews!

Communicating with Interviewees

• Most professional method of contact: email
• Expect at least a 1 week delay between emailing your interviewee and receiving a response
• Send a friendly follow-up email if the interviewee hasn’t responded for 1 week
• Responses are NOT guaranteed - if no response for some time, don’t take it personally
• Find a backup interviewee and reach out

• Make SURE you have done your research on your interviewee
• Position, background, experiences, achievements, etc.
• Use these to your advantage
• Ask your intervieweees questions that allow them to draw from their experiences - remember your interviewee is giving you expert testimonial

Communicating with Interviewees

• -

Interview Activity

1. Mentors: Give your mentees some background about your interests, major, extracurriculars, etc.
2. Mentees: Draft some questions for your mentees based on their background, and conduct a practice “interview” with them

To Do

Over the coming weeks…

1. Finalize an annotated list of ~5 sources for your project
2. Draft and send out emails to your prospective interviewees
3. Draft the questions you’d like to ask during your interview.