Announcements

FREE STUFF!!!
- Books are here
- DataCamp Access to Premium Content for 6 months
Final Project Reminders
- Canvas drop boxes
- Optional class feedback
  - 1-2 per week
  - first come first served
  - 5 minutes to talk about project (progress, problems, etc)
  - 5 minutes feedback from class & instructors
- Optional peer reviews
- DataCamp Alternative
  - Credit awarded for completion of assigned courses listed on DataCamp.com
  - 1 course: 10% credit
  - 2 courses: 20% credit
  - 3 courses: 30% credit
  - 4 courses: 50% credit
  - 5 courses: 70% credit
  - 6 courses: 80% credit
  - 7 courses: 85% credit
  - Note: Introduction to R course does NOT count toward Final Project alternative

Chapter 14: Key Ideas

Statistics is the area of science concerned with characterizing case-to-case variation and the collective properties of cases
It’s one thing to simply calculate quantities from data, but a central topic in statistics is the precision with which we can estimate those quantities
Quantifying uncertainty is perhaps the backbone of inferential statistics (e.g. confidence intervals, hypothesis tests)

it’s often useful to look at distributions of data graphically, and you were presented several useful ways to do so
- density plots (perhaps overlaid or faceted to highlight group comparisons)
- box plots (again, side-by-side or faceted to highlight group comparisons)
- violin plots (alterantive to box plots with greater detail in the density)
model functions help describe a relationship between an “input” (X, explanatory, independent) variable and an “output” (Y, response, dependent) variable
- smoothers show relationships between variables that bend with the data collectively
- linear functions are also useful in some contexts, but sometimes miss important features of the relationship between variables
confidence intervals (when estimating a single quantity) and confidence bands (when estimating model functions) are essential to communicate uncertainty
- small sample sizes result in a great deal of uncertainty
- we can be more confident in estimates produced by large sample sizes
- error bars, notched box plots, and confidence bands are all strategies to communicate uncertainty in a graph.

DC Ch 14 Exercises: 14.1, 14.3, 14.7, 14.10
DC chapter 16 reading quiz on Canvas
Activity: Statistics of Gene Expression
Choose final project data & submit on Canvas for approval (not graded; due 11/13 at 11:59pm)

Assignment is worth a total of 10 points.

[1 points] Turn in HTML with embedded .Rmd file (e.g. “DataComputing simple” template)
[5 points] Follow along with the activity in the book and reproduce the following figures as they appear in the book
- Figure A.6: Bar chart comparing TOP3A expression in different tissue types
- Figure A.7: TOP3A expression in the individual cells
- Figure A.8: TOP3A Bar chart with individual cells overlaid
- Figure A.9: TOP3A Dynamite plot (bar chart with error bars)
- Figure A.10: “Better than Dynamite” (TOP3A expression in the individual cells with overlaid confidence intervals by tissue type)
[1 point] recreate Figure A.11 showing 30 Probes with largest r-squared for expression level explained by tissue type
[2 points] Your Turn Choose one probe with high R^2. Plot expression versus tissue type (like Figure A.10)
[1 point] recreate Figures A.12 (overlaid density plots) & A.13 (comparison of highest R^2 from actual data and the Null)

If you follow along closely with the book instructions, you should be able to complete most of the activity
Among the “several suggestions for improving the graphic”, #3 will be most handy (calculating the mean and standard error)
You’ll need to combine elements of previous plots in order to produce Figure A.10
When you reproduce Figure A.10 for a new Probe in the Your Turn portion, you shouldn’t start from scratch. Find code you can copy and paste earlier in the assignment and modify it slightly to change the Probe of interest.