2014-2015 presentation series: Switch2R

• October 2014: R basics
• November 2014: Clean data with R
• December 2014 Analysis with R
• January 2015 Graphics with R (1/2)
• February 2015: Graphics with R (2/2)
• March 2015 Reproducible research with R (Dale)
• April 2015: GIS with R
• May 2015: Power and sample size
• June, July Aug. 2015: Summer break

Where to find code

on github (https://github.com/rnj0nes/Switch2R)

Updates from October 2014

• If you are a Windows user, you should be using Revolution R Open
  • It is much faster
  • Otherwise, no big difference from regular R
  • Integrates seamlessly with RStudio
  • A few weeks behind new R releases

• If you are a Mac user, don’t use Revolution R Open

  • Unless someone knows why you should

• If you are a Mac user, there was an issue with R, RStudio with the Yosemite Mac OS X 10.10 update

  • read more about it here

November 2014: Clean Data with R

• Use R to clean data

  • read data (review from last time)
  • make new variables
  • produce dataset documentation
  • label variables and values (sjPlot)

Things to remember about R

• There are about 57 different ways to do the same thing
• There is undoubtedly a better way to do the things I am going to show you
• When you find out please share

Things to remember about data in R

• R natively does not support variable and value labels
• You can load multiple data sets in R at once

• Data can be

  • in data frames
  • in matrix
  • attached (but don’t do that)
  • in a list

What are clean data?

• Clean data are tidy data
• Tidy data are not messy data

• Tidy data have the following characteristics

  • 1 variable : 1 column
  • 1 observation : 1 row
  • Each type of observational unit forms a table, or,
  • data are only stored in one place (you may need different tables)

Example from last month

Atkins, D. C. (2005). Using multilevel models to analyze couple and family treatment data: Basic and advanced issues. Journal of Family Psychology, 19, 98-110.

Data from UCLA IDRE

Article: findit at Brown full text (http://goo.gl/D6m2ID)

R Session

• Decide where you will do your work
• Set the working folder
• Download data
• Write (and save) code

What I do

• I work in folder c:/work/project-name-here

work <- "c:/work/shows/switch2r"
setwd(work)
getwd()

## [1] "c:/work/shows/switch2r"

Load some packages

(note: order matters)

# install.packages("lubridate")
# install.packages("psych")
# install.packages("gmodels")
# install.packages("Hmisc")
# install.packages("sjPlot")
require(lubridate) # easy handling of dates and times
require(psych)    # using the scoreItems command 
require(Hmisc)    # using the describe command
require(sjPlot)   # tools for reading SPSS formatted data
require(gmodels)  # using the CrossTable command

Read in the data set from the web

# Read data from UCLA Web site
url <- "http://www.ats.ucla.edu/stat/paperexamples/atkins_mlm/Atkins_JFP_data.txt"
data <- read.csv(url, sep="\t", header=TRUE)
# I like all lowercase variable names
names(data)<- tolower(names(data)) 
# show the first 15 lines
head(data, 15)

##    id sex therapy time       das pilot miss m.ind
## 1   1   0    -0.5    0  94.51204     1    0     1
## 2   1   0    -0.5   13  87.53364     1    0     1
## 3   1   0    -0.5   26  81.46659     1    1     1
## 4   1   0    -0.5   35  83.44614     1    1     1
## 5   1   1    -0.5    0  81.27981     1    0     1
## 6   1   1    -0.5   13  68.80343     1    0     1
## 7   1   1    -0.5   26  71.16971     1    1     1
## 8   1   1    -0.5   35  76.88723     1    1     1
## 9   2   0     0.5    0  79.53347     1    0     0
## 10  2   0     0.5   13  98.75209     1    0     0
## 11  2   0     0.5   26 100.15992     1    0     0
## 12  2   0     0.5   35 127.55920     1    0     0
## 13  2   1     0.5    0 106.98098     1    0     0
## 14  2   1     0.5   13 124.13277     1    0     0
## 15  2   1     0.5   26 143.09363     1    0     0

Are these data tidy?

• Discuss

Make some variables

The data file has a variable therapy that is coded -.5/+.5.

Let’s say I’d like to have a version of that variable that was coded 1/2.

I’ll call it tx

data$tx <- 1 # New variable tx in data.frame data = 1 (by default)
data$tx[which(data$therapy==0.5)] <- 2 # if therapy == 0.5 data$tx = 2
str(data[,c("tx","therapy")]) # show the characteristics of two variables

## 'data.frame':    1072 obs. of  2 variables:
##  $ tx     : num  1 1 1 1 1 1 1 1 2 2 ...
##  $ therapy: num  -0.5 -0.5 -0.5 -0.5 -0.5 -0.5 -0.5 -0.5 0.5 0.5 ...

# Now check
table(data$tx, data$therapy) # a crosstab of tx and therapy in data.frame data

##    
##     -0.5 0.5
##   1  536   0
##   2    0 536

A nicer looking cross-tab

Using CrossTable from the gmodels package

CrossTable(data$tx, data$therapy, 
  missing.include=TRUE, 
  prop.r=FALSE, 
  prop.c=FALSE, 
  prop.t=FALSE, 
  prop.chisq=FALSE)

## 
##  
##    Cell Contents
## |-------------------------|
## |                       N |
## |-------------------------|
## 
##  
## Total Observations in Table:  1072 
## 
##  
##              | data$therapy 
##      data$tx |      -0.5 |       0.5 | Row Total | 
## -------------|-----------|-----------|-----------|
##            1 |       536 |         0 |       536 | 
## -------------|-----------|-----------|-----------|
##            2 |         0 |       536 |       536 | 
## -------------|-----------|-----------|-----------|
## Column Total |       536 |       536 |      1072 | 
## -------------|-----------|-----------|-----------|
## 
## 

But that is so cumbersome to code…

Welcome to R

tab <- function(r,c) {
  CrossTable(r, c, 
    missing.include=TRUE, prop.r=FALSE, prop.c=FALSE, prop.t=FALSE, 
    prop.chisq=FALSE)
}

Now a nice cross-tab

tab(data$tx,data$therapy)

## 
##  
##    Cell Contents
## |-------------------------|
## |                       N |
## |-------------------------|
## 
##  
## Total Observations in Table:  1072 
## 
##  
##              | c 
##            r |      -0.5 |       0.5 | Row Total | 
## -------------|-----------|-----------|-----------|
##            1 |       536 |         0 |       536 | 
## -------------|-----------|-----------|-----------|
##            2 |         0 |       536 |       536 | 
## -------------|-----------|-----------|-----------|
## Column Total |       536 |       536 |      1072 | 
## -------------|-----------|-----------|-----------|
## 
## 

And anyway the coding looks right. But notice the variable names are now the not-so-helpful “r” and “c”. So whatever.

Watch me do this with Rcmdr

Link to video

Sex

Another example. I’ll code a variable male from the original sex

data$male <- 0 # initialize to 0
data$male[which(data$sex!=1)] <- 1 # sex is 0:Husband 1:Wife
tab(data$male,data$sex)

## 
##  
##    Cell Contents
## |-------------------------|
## |                       N |
## |-------------------------|
## 
##  
## Total Observations in Table:  1072 
## 
##  
##              | c 
##            r |         0 |         1 | Row Total | 
## -------------|-----------|-----------|-----------|
##            0 |         0 |       536 |       536 | 
## -------------|-----------|-----------|-----------|
##            1 |       536 |         0 |       536 | 
## -------------|-----------|-----------|-----------|
## Column Total |       536 |       536 |      1072 | 
## -------------|-----------|-----------|-----------|
## 
## 

About indicator variables

male is a binary indicator (0/1). When coding binary indicators with the values (0/1), always code so that the name of the variable always matches the label for value 1. For example,

R sex	Categorical Variable	Binary indicator	Binary indicator
is really	Gender codes	Female	Male
a man	1 = male	0	1
a woman	2 = female	1	0

Dates

The example data set does not have any date information.

Public data rarely will.

But you will surely have to deal with dates.

So let’s pretend it has dates.

To pretend it has dates, we will generate some date data. So now you get to see how we generate data with R.

set.seed(3481)
data$year <- sample(1:3,nrow(data),replace=T)+2008
data$day <-  sample(1:27,nrow(data),replace=T)
data$month <- round(runif(nrow(data),0.51,12.49))

The first line says: make a new variable year in data frame data. Assign to it values, sample from the range 1 to 3, as many times as there are rows in the data frame named data. Sample with replacement. Then add 2008.

The third line says: make a new variable month in data frame data. Assign to it a random numbers (as many as there are rows in the data frame) drawn from a uniform distribution ranging from 0.51 to 12.49. Oh but round it to the nearest whole number too.

Display the structure of the day month year variables

str(data[,c("day","month","year")])

## 'data.frame':    1072 obs. of  3 variables:
##  $ day  : int  5 1 27 16 3 18 19 24 3 6 ...
##  $ month: num  2 4 3 10 5 5 5 11 3 5 ...
##  $ year : num  2011 2010 2010 2010 2011 ...

Tabulate the day month year variables

table(data$day)

## 
##  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 
## 41 37 45 43 37 49 34 36 44 31 32 33 40 37 42 33 44 35 47 40 49 35 37 47 42 
## 26 27 
## 42 40

table(data$month)

## 
##   1   2   3   4   5   6   7   8   9  10  11  12 
##  75  83  87  72 102  93  84  83  72 106 105 110

table(data$year)

## 
## 2009 2010 2011 
##  363  363  346

Dates using lubridate

(luridate vignette)[http://cran.r-project.org/web/packages/lubridate/vignettes/lubridate.html]

Some interesting things about dates

the.time.is.now <- now()
the.time.is.now.numeric <- as.numeric(now())
the.time.is.now

## [1] "2014-11-12 15:54:03 EST"

the.time.is.now.numeric

## [1] 1415825644

as.numeric(mdy("1/1/1970"))

## [1] 0

as.numeric(mdy("1/2/1970"))

## [1] 86400

60*60*24

## [1] 86400

str(the.time.is.now)

##  POSIXct[1:1], format: "2014-11-12 15:54:03"

Making a datetime variable

data$year <- as.character(data$year)
data$month <- as.character(data$month)
data$day <- as.character(data$day)
data$date <- mdy(paste0(data$month,"/",data$day,"/",data$year))
head(data[,c("year","month","day","date")])

##   year month day       date
## 1 2011     2   5 2011-02-05
## 2 2010     4   1 2010-04-01
## 3 2010     3  27 2010-03-27
## 4 2010    10  16 2010-10-16
## 5 2011     5   3 2011-05-03
## 6 2009     5  18 2009-05-18

New Example: National Comorbidity Study Replication (NCS-R 2002-2003)

Data from ICPSR (public use)

I can redistribute to Brown faculty, students, and staff (authorized users)

If you are not Brown faculty, students, and staff, see about getting access to the NCS-R public use files at ICPSR

NCS-R data

I am pulling the NCS-R data as archived along with the Collaborative Psychiatric Epidemiology Surveys.

CPES Website

Download the SPSS data file

Let’s get set

work <- "c:/work/shows/switch2r" # my working folder you edit here
setwd(work) # change to it
getwd() # check it

## [1] "c:/work/shows/switch2r"

I’ve downloaded the SPSS file to a subfolder here (NCSR)

dir("./NCSR")

## [1] "20240-0002-Codebook.pdf"      "20240-0002-Data.sav"         
## [3] "20240-0002-Data.txt"          "20240-0002-Documentation.pdf"
## [5] "20240-0002-Setup.sas"

We’ll use the sjPlot package to read in the SPSS data

and generate a nice data dictionary

spssdata <- sji.SPSS("./NCSR/20240-0002-Data.sav")
# sji.viewSPSS(spssdata) # opens in the Rstudio Viewer Panel
# Or in a browser window if you're running the R Console

Let’s save that codebook

#sji.viewSPSS(spssdata, file="codebook.html", showFreq = TRUE, useViewer = FALSE)
codebookpath <- file.path(work,"codebook.html")

The codebook

But that link only works on your (i.e., my) local computer

Viewers of this presentation can see the codebook I generated on my dropbox

Let’s make a summated rating scale score

Psychotic experiences

• PS1A Ever see vision that others couldn’t see
• PS1B Ever hear voices others couldn’t hear
• PS1C Ever have mind control experience
• PS1D Ever feel mind taken over by strange forces
• PS1E Ever exp communication attempts from strange forces
• PS1F Unjust plot to harm you/have people follow-nobody believe

Each coded 1=Yes, 5=No

Look at the data

PSitems <- c("PS1A", "PS1B", "PS1C", "PS1D", "PS1E", "PS1F")
describe(spssdata[PSitems])

## spssdata[PSitems] 
## 
##  6  Variables      9282  Observations
## ---------------------------------------------------------------------------
## PS1A 
##       n missing  unique    Info    Mean 
##    2349    6933       2    0.25    4.63 
## 
## 1 (217, 9%), 5 (2132, 91%) 
## ---------------------------------------------------------------------------
## PS1B 
##       n missing  unique    Info    Mean 
##    2349    6933       2    0.17   4.753 
## 
## 1 (145, 6%), 5 (2204, 94%) 
## ---------------------------------------------------------------------------
## PS1C 
##       n missing  unique    Info    Mean 
##    2352    6930       2    0.02   4.968 
## 
## 1 (19, 1%), 5 (2333, 99%) 
## ---------------------------------------------------------------------------
## PS1D 
##       n missing  unique    Info    Mean 
##    2353    6929       2    0.01   4.981 
## 
## 1 (11, 0%), 5 (2342, 100%) 
## ---------------------------------------------------------------------------
## PS1E 
##       n missing  unique    Info    Mean 
##    2352    6930       2    0.04   4.951 
## 
## 1 (29, 1%), 5 (2323, 99%) 
## ---------------------------------------------------------------------------
## PS1F 
##       n missing  unique    Info    Mean 
##    2352    6930       2    0.05   4.937 
## 
## 1 (37, 2%), 5 (2315, 98%) 
## ---------------------------------------------------------------------------

Looks like a skip pattern

Kessler says…

• A total of 9282 respondents participated (2001-2003)
• All respondents completed a Part I diagnostic interview (WHO-CIDI)
• A probability sample of 5692 also received Part II (additional disorders)
• A random sub-sample of Part II respondents (n = 2322) was administered the NAP screen
• NAP (Non-affective psychosis)
• Our N=2349-2353 is pretty close to that

Kessler RC, Birnbaum H, Demler O, Falloon IR, Gagnon E, Guyer M, Howes MJ, Kendler KS, Shi L, Walters E. The prevalence and correlates of nonaffective psychosis in the National Comorbidity Survey Replication (NCS-R). Biol Psychiatry. 2005;58(8):668-76.

Right ways and wrong ways to make a sum score

Remember

table(spssdata$PS1A)

## 
##    1    5 
##  217 2132

Right ways and wrong ways to make a sum score

This works

spssdata$count.ok <- (spssdata$PS1A==1)+
  (spssdata$PS1B==1)+(spssdata$PS1C==1)+
  (spssdata$PS1D==1)+(spssdata$PS1E==1)+
  (spssdata$PS1F==1)

Right ways and wrong ways to make a sum score

This does not

spssdata$count.notok <- (spssdata$PS1A==1)
  +(spssdata$PS1B==1)
  +(spssdata$PS1C==1)
  +(spssdata$PS1D==1)
  +(spssdata$PS1E==1)
  +(spssdata$PS1F==1)

Right ways and wrong ways to make a sum score

See

spssdata$count.ok <- (spssdata$PS1A==1)+
  (spssdata$PS1B==1)+(spssdata$PS1C==1)+
  (spssdata$PS1D==1)+(spssdata$PS1E==1)+
  (spssdata$PS1F==1)
summary(spssdata$count.ok)

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max.    NA's 
##   0.000   0.000   0.000   0.194   0.000   6.000    6938

Right ways and wrong ways to make a sum score

Let’s check

spssdata$check <- paste0(as.character(spssdata$PS1A),
  as.character(spssdata$PS1B),as.character(spssdata$PS1C),
  as.character(spssdata$PS1D),as.character(spssdata$PS1E),
  as.character(spssdata$PS1F))

Right ways and wrong ways to make a sum score

Let’s check

head(spssdata[which(spssdata$count==0),c("check","count.ok")])

##     check count.ok
## 8  555555        0
## 12 555555        0
## 18 555555        0
## 24 555555        0
## 25 555555        0
## 26 555555        0

Right ways and wrong ways to make a sum score

Let’s check

head(spssdata[which(spssdata$count==1),c("check","count.ok")])

##      check count.ok
## 17  155555        1
## 19  155555        1
## 212 155555        1
## 321 515555        1
## 378 155555        1
## 381 555551        1

Right ways and wrong ways to make a sum score

Let’s check

head(spssdata[which(spssdata$count==2),c("check","count.ok")])

##      check count.ok
## 313 115555        2
## 367 515551        2
## 453 115555        2
## 501 155551        2
## 660 115555        2
## 721 155515        2

Right ways and wrong ways to make a sum score

Let’s check

head(spssdata[which(spssdata$count==3),c("check","count.ok")])

##       check count.ok
## 1176 115155        3
## 2254 115551        3
## 3084 115155        3
## 3620 111555        3
## 3658 115551        3
## 4115 111555        3

Right ways and wrong ways to make a sum score

Let’s check

head(spssdata[which(spssdata$count==4),c("check","count.ok")])

##       check count.ok
## 38   151115        4
## 3793 111551        4
## 4426 115115        4
## 6260 115511        4
## 6299 115511        4
## 7040 111515        4

Right ways and wrong ways to make a sum score

Let’s check

head(spssdata[which(spssdata$count==5),c("check","count.ok")])

##       check count.ok
## 6721 111511        5

Right ways and wrong ways to make a sum score

Let’s check

head(spssdata[which(spssdata$count==6),c("check","count.ok")])

##       check count.ok
## 1453 111111        6
## 4460 111111        6

Using scoreItems from package psych

psitems <- c("ps1a.i", "ps1b.i", "ps1c.i", "ps1d.i", "ps1e.i", "ps1f.i")
# Make indicator versions of response items
spssdata$ps1a.i <- as.numeric(spssdata$PS1A==1)
spssdata$ps1b.i <- as.numeric(spssdata$PS1B==1)
spssdata$ps1c.i <- as.numeric(spssdata$PS1C==1)
spssdata$ps1d.i <- as.numeric(spssdata$PS1D==1)
spssdata$ps1e.i <- as.numeric(spssdata$PS1E==1)
spssdata$ps1f.i <- as.numeric(spssdata$PS1F==1)

Using scoreItems from package psych

key <- c(1,1,1,1,1,1) # the "right" answers
results <- scoreItems(
  items = spssdata[psitems],
  keys = key, 
  missing = TRUE , impute = "none"  # person's non-missing mean
  #                                   response used for missing
  )

Display scoreItem results

results

## Call: scoreItems(keys = key, items = spssdata[psitems], missing = TRUE, 
##     impute = "none")
## 
## (Standardized) Alpha:
##         A1
## alpha 0.53
## 
## Standard errors of unstandardized Alpha:
##         [,1]
## ASE   0.0094
## 
## Standardized Alpha of observed scales:
##                [,1]
## alpha.observed 0.53
## 
## Average item correlation:
##           [,1]
## average.r 0.16
## 
##  Guttman 6* reliability: 
##          [,1]
## Lambda.6 0.52
## 
## Signal/Noise based upon av.r : 
##              [,1]
## Signal/Noise  1.1
## 
## Scale intercorrelations corrected for attenuation 
##  raw correlations below the diagonal, alpha on the diagonal 
##  corrected correlations above the diagonal:
## 
## Note that these are the correlations of the complete scales based on the correlation matrix,
##  not the observed scales based on the raw items.
##      [,1]
## [1,] 0.53
## 
##  In order to see the item by scale loadings and frequency counts of the data
##  print with the short option = FALSE

Display the structure of the results

str(results)

## List of 17
##  $ scores        : num [1:9282, 1] NaN NaN NaN NaN NaN NaN NaN 0 NaN NaN ...
##   ..- attr(*, "dimnames")=List of 2
##   .. ..$ : NULL
##   .. ..$ : chr "A1"
##  $ missing       : num [1:9282, 1] 6 6 6 6 6 6 6 0 6 6 ...
##  $ alpha         : num [1, 1] 0.527
##   ..- attr(*, "dimnames")=List of 2
##   .. ..$ : chr "alpha"
##   .. ..$ : chr "A1"
##  $ av.r          : num [1, 1] 0.157
##   ..- attr(*, "dimnames")=List of 2
##   .. ..$ : chr "average.r"
##   .. ..$ : NULL
##  $ sn            : num [1, 1] 1.11
##   ..- attr(*, "dimnames")=List of 2
##   .. ..$ : chr "Signal/Noise"
##   .. ..$ : NULL
##  $ n.items       : num 6
##  $ item.cor      : num [1:6, 1] 0.755 0.726 0.445 0.359 0.517 ...
##   ..- attr(*, "dimnames")=List of 2
##   .. ..$ : chr [1:6] "ps1a.i" "ps1b.i" "ps1c.i" "ps1d.i" ...
##   .. ..$ : chr "A1"
##  $ cor           : num [1, 1] 1
##  $ corrected     : num [1, 1] 0.527
##  $ G6            : num [1, 1] 0.523
##   ..- attr(*, "dimnames")=List of 2
##   .. ..$ : chr "Lambda.6"
##   .. ..$ : NULL
##  $ item.corrected: num [1:6, 1] 0.44 0.521 0.427 0.353 0.498 ...
##   ..- attr(*, "dimnames")=List of 2
##   .. ..$ : chr [1:6] "ps1a.i" "ps1b.i" "ps1c.i" "ps1d.i" ...
##   .. ..$ : chr "A1"
##  $ response.freq : num [1:6, 1:3] 0.908 0.938 0.992 0.995 0.988 ...
##   ..- attr(*, "dimnames")=List of 2
##   .. ..$ : chr [1:6] "ps1a.i" "ps1b.i" "ps1c.i" "ps1d.i" ...
##   .. ..$ : chr [1:3] "0" "1" "miss"
##  $ raw           : logi FALSE
##  $ alpha.ob      : num [1, 1] 0.527
##   ..- attr(*, "dimnames")=List of 2
##   .. ..$ : chr "alpha.observed"
##   .. ..$ : NULL
##  $ num.ob.item   : num 6
##  $ ase           : num [1, 1] 0.00941
##  $ Call          : language scoreItems(keys = key, items = spssdata[psitems], missing = TRUE,      impute = "none")
##  - attr(*, "class")= chr [1:2] "psych" "score.items"

Using scoreItems from package psych

summary(results$score)

##        A1       
##  Min.   :0.000  
##  1st Qu.:0.000  
##  Median :0.000  
##  Mean   :0.032  
##  3rd Qu.:0.000  
##  Max.   :1.000  
##  NA's   :6929

Using scoreItems from package psych

results$pscount <- 6*results$score
summary(results$pscount)

##        A1       
##  Min.   :0.000  
##  1st Qu.:0.000  
##  Median :0.000  
##  Mean   :0.195  
##  3rd Qu.:0.000  
##  Max.   :6.000  
##  NA's   :6929

But

How do we get results$pscount back into the working data set?

spssdata$pscount.master <- scoreItems(items = spssdata[psitems],
  keys = key, missing = TRUE , impute = "none")[[1]]*6
summary(spssdata$pscount.master)

##        A1       
##  Min.   :0.000  
##  1st Qu.:0.000  
##  Median :0.000  
##  Mean   :0.195  
##  3rd Qu.:0.000  
##  Max.   :6.000  
##  NA's   :6929

Display the internal consistency reliability coefficient

results$alpha

##              A1
## alpha 0.5271367

Save alpha to two decimal places

alpha <- round(results$alpha,2)

My markdown code

The internal consistency reliability coefficient for the sum of the
psychotic experiences scale is 'r alpha'.

The internal consistency reliability coefficient for the sum of the psychotic experiences scale is 0.53.

Display the structure and report the item response frequencies

str(results$response.freq)

##  num [1:6, 1:3] 0.908 0.938 0.992 0.995 0.988 ...
##  - attr(*, "dimnames")=List of 2
##   ..$ : chr [1:6] "ps1a.i" "ps1b.i" "ps1c.i" "ps1d.i" ...
##   ..$ : chr [1:3] "0" "1" "miss"

results$response.freq

##                0           1      miss
## ps1a.i 0.9076203 0.092379736 0.7469295
## ps1b.i 0.9382716 0.061728395 0.7469295
## ps1c.i 0.9919218 0.008078231 0.7466063
## ps1d.i 0.9953251 0.004674883 0.7464986
## ps1e.i 0.9876701 0.012329932 0.7466063
## ps1f.i 0.9842687 0.015731293 0.7466063

Review of things we accomplished today

• Big ideas
• Packages
• Commands and functions
• Skills and tasks

Big ideas

• using public data
• tidy data
• documentation
• seeds

Packages

• lubridate
• gmodels (CrossTab)
• sjPlot (sji.SPSS and others)
• psych

Commands and functions

• setwd, getwd
• CrossTable (gModels)
• sample
• runif
• names
• tolower
• as.numeric, as.character
• head
• str
• file.path
• paste0
• scoreItems (psych)

Skills and tasks

• install packages
• load packages
• set the working directory
• relative paths
• read data from the internet
• read data from SPSS/SAV file
• refer to specific variables
• refer to specific observations
• made a cross tab
• made new variables
• generate random data
• define a new function
• working with dates

Next meeting

Analysis with R

Use R to do multivariable analysis

• Linear regression
• Logistic regression
• Survival analysis
• Repeated measures mixed models

Monday, December 8, 2014, 2-3pm.