Welcome/Course Intro

Univariate Statistics and Methodology using R

Martin Corley

Psychology, PPLS

University of Edinburgh

Welcome to Univariate Statistics and Methodology Using R

Univariate Statistics…

heights

# A tibble: 102 × 4
   gender    HEIGHT submitted           Token                           
   <chr>      <dbl> <chr>               <chr>                           
 1 Female       173 02/10/2019 08:03:18 fo8qg8m4w53yil2nzufo8qxzpjkgzj0l
 2 Female       159 02/10/2019 08:03:19 5jjo15xpzs23gdt5esks7jd7ol3uug8p
 3 Female       164 02/10/2019 08:03:20 q2i04mqiaigqv6w82skh8orq2i04mujp
 4 Male         187 02/10/2019 08:03:25 2n3x6kpuqjfhv3x2n3x63g7sqs3oshym
 5 Female       183 02/10/2019 08:03:26 xsi090b2wfjb61xeryhxsi0q867xls2i
 6 NonBinary    173 02/10/2019 08:03:27 feg5s1kerz3dw5090qfeg5s1keg6x08w
 7 Female       168 02/10/2019 08:03:27 ph3e74tai0l3urg26ph34qfsavvxkq5g
 8 Female       170 02/10/2019 08:03:30 dbzsfmoir3mkwtlvi7udbz0iw53p0r83
 9 Male         175 02/10/2019 08:03:36 qt2vujoe9ka2867o28sqt2vujm9cz1h9
10 Male         165 02/10/2019 08:03:36 cvnzmtez3ej4oa54jfbpcvnzmqcgr3jx
# ℹ 92 more rows

Univariate Statistics…

heights

# A tibble: 102 × 4
   gender    HEIGHT submitted           Token                           
   <chr>      <dbl> <chr>               <chr>                           
 1 Female       173 02/10/2019 08:03:18 fo8qg8m4w53yil2nzufo8qxzpjkgzj0l
 2 Female       159 02/10/2019 08:03:19 5jjo15xpzs23gdt5esks7jd7ol3uug8p
 3 Female       164 02/10/2019 08:03:20 q2i04mqiaigqv6w82skh8orq2i04mujp
 4 Male         187 02/10/2019 08:03:25 2n3x6kpuqjfhv3x2n3x63g7sqs3oshym
 5 Female       183 02/10/2019 08:03:26 xsi090b2wfjb61xeryhxsi0q867xls2i
 6 NonBinary    173 02/10/2019 08:03:27 feg5s1kerz3dw5090qfeg5s1keg6x08w
 7 Female       168 02/10/2019 08:03:27 ph3e74tai0l3urg26ph34qfsavvxkq5g
 8 Female       170 02/10/2019 08:03:30 dbzsfmoir3mkwtlvi7udbz0iw53p0r83
 9 Male         175 02/10/2019 08:03:36 qt2vujoe9ka2867o28sqt2vujm9cz1h9
10 Male         165 02/10/2019 08:03:36 cvnzmtez3ej4oa54jfbpcvnzmqcgr3jx
# ℹ 92 more rows

• one row represents one set of observations

• typically in Psychology, this might be the identity of the person being measured
• example here - we have different peoples heights.
• the variation comes from all these different people we’ve measured
• as statisticians, we’re interested in examining whether there are systematic differences, beyond the individual variation
• for example, height predicted by gender?

…and Methodology…

• how do we measure height?
• how do we determine gender?
• how do we collect data in a way that is generalisable?

…Using R…

heights |>
    ggplot(aes(x = HEIGHT,
        colour = gender,
        fill = gender)) +
    geom_density(size = 2,
        alpha = 0.4)

…to Form Conclusions

heights_t <- heights |>
  filter(gender %in% c("Male", "Female"))

t.test(heights_t$HEIGHT ~ heights_t$gender)

    Welch Two Sample t-test

data:  heights_t$HEIGHT by heights_t$gender
t = -9, df = 29, p-value = 0.0000000006
alternative hypothesis: true difference in means between group Female and group Male is not equal to 0
95 percent confidence interval:
 -16.26 -10.25
sample estimates:
mean in group Female   mean in group Male 
               165.7                179.0 

How the Course Works

What We Will Learn

statistics

• foundations of Null Hypothesis Significance Testing
• probability
• samples and distributions
• the normal and the binomial distributions
• testing for significance
  • \(F\)-ratio, \(\chi^2\), \(t\)-test, others
• the linear model
• multiple linear regression
• assumptions, models, model criticism
• logit regression (generalized linear model)

R

• basic programming
• using libraries
• using an IDE (RStudio)
• data types
  • data manipulation
  • visualisation (graphs)
• functions
• running statisical models
• RMarkdown
  • literate programming
  • document creation

• course covers a whole bunch of statistics.

• based on particular branch - NHST, which is the dominant approach in most fields, where we assume that the thing we are observing is governed by chance, and we work out the probability of what we have observed occuring by chance.
  

• and if that’s low, then we assume that other factors are at play.
  

• we’ll be needing to cover things about probability, samples and populations, and tests of significance.

• the second block, starting week 6, starts to shift our focus to the linear model, which is a lovely way of thinking about analyses that provides the building blocks of many many complicated analyses. lots of the tests and models you might read about are either special cases of, or extensions of, the linear model.
  

Shape of the Course

Lectures
often include live coding

Readings/Walkthroughs
you’re encouraged to work along with these

Labs (Exercises)
work in groups, with help on hand from a team of tutors

Discussion Forums and Support
via learn page

Assessment

Lectures & Readings

• broadly, about concepts

  • statistics

  • coding

Exercises

• broadly, how to

  • coding

  • data manipulation

  • statistics

• lots of hints, links to readings

solutions will be available at the end of each week

Labs

• a time and place to work on the exercises

• you will be working in groups

• a team of tutors will be there to help

• labs are the best place to get to grips with R and statistics
• you are expected to attend

• statistics, and especially R, is a very practical skill.

• we can’t teach people to swim by just giving lectures on how to swim.

Discussions

• piazza discussion forums for the course on Learn

  • ask questions, share experiences, talk to the course team

  • post anonymously if preferred

  • an important way to keep in touch

Support

we are here to help you

• lectures: feel free to ask questions

• labs: ask the tutors (they want to help!)

• piazza discussion forums: any time

• office hours: see Learn page for details

Course Quizzes (35%)

• 1 practice quiz and 9 assessed quizzes
• best 7 of the 9 assessed quizzes will count towards the final grade
• quizzes each have approximately 10 questions
• for each quiz, one attempt which must be completed within 60 min

released each Friday at 17:00

due the following Friday at 17:00

quizzes should be taken individually

Group Project (65%)

• check a dataset for consistency

• explore hypotheses about how variables are related

  • conduct and interpret appropriate statistical tests

• produce suitable graphics

• present workings and conclusions in a report

released Thursday 9 November

due Thursday 14 December at 16:00

each lab group should produce one report (with code)

Tips for Survival

• active engagement!

• use the piazza forums and other forms of support

• keep on top of the coursework

• remember that some things will feel difficult at first

  • what’s hard for you may be easy for others
  • what’s easy for you may be hard for others

• most importantly, don’t give up

attend the lectures, and don’t be afraid to ask any questions.

if you don’t understand something one of us says, then my guess is that either we made a mistake, or we didn’t explain ourselves very well.

there’s a bit of an imposter syndrome problem in statistics, and it’s very hard to tackle.

important to remember why we’re here, and that is to learn.

Tips for Improving the Course

• we work constantly to improve this course, but fixing some things may well break others

• please feed back to us!

  • during labs/after lectures
  • via email, via forums, anonymously via “Have Your Say” on Learn

• please bear with us, we are doing our best

• any good course is a conversation between teachers and learners

  • the course will only work well with your commitment and input

Get Going

1. Help Us Get to Know You

• log into Learn, navigate to the USMR course page

• look for the Piazza Forum button

  • it should automatically enroll you on piazza

• post a little introduction of yourself

  Suggested content:

  • pictures of pets/animals very welcome.

  • answer a randomly generated question from https://learnhip.com/randomq/

2. Get the Software

Option A

RStudio Server

• https://rstudio.ppls.ed.ac.uk,
  no installation required
• accessed via a web browser (requires internet)

Option B

Install R and RStudio

• instructions at https://edin.ac/3B0oi5A
  
• process is a bit more involved

our recommendation

• use Rstudio Server for now

• install R and RStudio over Xmas, in time for Semester 2

• The PPLS Rstudio Server will undergo regular maintenance on the first Sunday of every month, so there may be a short interruption to the service on these days
  
• Our agreement with RStudio allows us the use of RStudio Server for teaching purposes only. Please do not use the RStudio Server for your dissertation
  

3. Fill in the Survey

• find the survey at https://edin.ac/3Eth1ML

• takes 10–15 minutes to complete

• provides real data for the examples and exercises

End