• statistics on a Mac

• 80% preparing data, 20% complaining about preparing data

• 90% Googling and 10% pasting from Stack Overflow

If at first you don’t succeed,
call it version 1.0!

• Math and statistics

• Computer science

• Knowledge about the system

• Communication

1. Define the question of interest
   
2. Get the data
   
3. Clean the data
   
4. Explore the data
   
5. Fit statistical models
   
6. Communicate the results

1. Define the question of interest
   
2. Get the data
   
3. Clean the data
   
4. Explore the data
   
5. Fit statistical models
   
6. Communicate the results

1. Define the question of interest
   
2. Get the data
   
3. Clean the data
   
4. Explore the data
   
5. Fit statistical models
   
6. Communicate the results

“Good data science is distinguished from bad data science primarily by a repeatable, thoughtful, skeptical application of an analytic process to data in order to arrive at supportable conclusions.”

– Jeff Leek, Vice President and Chief Data Officer, Fred Hutch

Summarizes the information in a single data set without further interpretation

ex: the US Census

Searches for trends, correlations, etc among multiple variables

ex: “data mining”

Generalizes to the population from a sample

ex: wearing a mask reduces the spread of COVID

Uses a subset of the data to parameterize a model and predicts out-of-sample data

ex: using weather to predict recreational fishing effort

Seeks to understand the average magnitude and direction of a response

ex: experimental analysis of temperature effects on oyster growth

Seeks to understand the specific magnitude and direction of relationships

ex: engineering studies

Although you may have spent weeks/months on an analysis, most people only want the bottom line

The key is weaving everything into a nice story to achieve both trust and belief

Do you accept the analysis per se?

Were the data analyzed properly and thoroughly?

Trust is specific to the analysis/analyst (internal)

Do you believe the analysis?

Belief is more broadly related to previous work and other factors (external)

1. The things you did and presented

2. The things you did but did not present

3. The things you did not do

The trick is to strike a balance to achieve trust and belief

• Thinking about data separately from research

• Increasing efficiency & reproducibility

• Facilitating collaboration with others

• Finding answers from the community

• Communicating our workflows & results

• Specific options for cleaning & tidying data (eg, tidyverse)

• Specific analytical techniques (eg, machine learning)

• Specific plotting techniques (eg, ggplot)

• Use lots of online resources

• Use open/live coding

• Rely on our community for help