Package 'walrus'

Walrus

Description

A toolbox of common robust statistical tests, including robust descriptives, robust t-tests, and robust ANOVA. It is also available as a module for 'jamovi' (see www.jamovi.org for more information). Walrus is based on the WRS2 package by Patrick Mair, which is in turn based on the scripts and work of Rand Wilcox. These analyses are described in depth in the book Introduction to Robust Estimation & Hypothesis Testing.

Details

Box & Violin Plots				rplots()
Robust Descriptives				rdesc()
Robust Independent Samples T-Test				rttestIS()
Robust Paired Samples T-Test				rttestIS()
Robust ANOVA				ranova()

Ravi:
    "Should we create a logo for walrus?"

Jonathon:
    "Yeah, I guess. Maybe a walrus, or a
    skewed distribution?
    Bonus points if it somehow contains both."

Ravi gets bonus points

See Also

Useful links:

• https://github.com/jamovi/walrus

• Report bugs at https://github.com/jamovi/walrus/issues

---

Robust ANOVA

Description

Robust Analysis of Variance

Usage

ranova(
  data,
  dep,
  factors = NULL,
  method = "trim",
  ph = FALSE,
  tr = 0.2,
  est = "mom",
  nboot = 599,
  dist = "proj"
)

Arguments

data	the data as a data frame
dep	a string naming the dependent variable from data; the variable must be numeric
factors	a vector of strings naming the fixed factors from data
method	'median', 'trim' (default) or 'boot'; the method to use, median, trimmed means, or bootstrapped
ph	TRUE or FALSE (default), provide post hoc tests
tr	a number between 0 and 0.5, (default: 0.2), the proportion of measurements to trim from each end, when using the trim and bootstrap methods
est	'onestep', 'mom' (default) or 'median', the M-estimator to use; One-step, Modified one-step or Median respectively
nboot	a number (default: 599) specifying the number of bootstrap samples to use when using the bootstrap method
dist	'maha' or 'proj' (default), whether to use Mahalanobis or Projection distances respectively

Value

A results object containing:

results$main					the table of ANOVA results
results$phs					the table of posthoc tests

Tables can be converted to data frames with asDF or as.data.frame. For example:

results$main$asDF

as.data.frame(results$main)

Examples

data('goggles', package='WRS2')

ranova(goggles,
       dep = 'attractiveness',
       factors = c('gender', 'alcohol'),
       ph = TRUE)

#
#  ROBUST ANOVA
#
#  Robust ANOVA
#  ------------------------------------
#                      Q        p
#  ------------------------------------
#    gender             1.67    0.209
#    alcohol           48.28    0.001
#    gender:alcohol    26.26    0.001
#  ------------------------------------
#   Note. Method of trimmed means,
#   trim level 0.2
#
#
#  POST HOC TESTS
#
#  Post Hoc Tests - gender
#  --------------------------------------------------------
#                       psi-hat   p        Lower    Upper
#  --------------------------------------------------------
#    Female    Male     10.0      0.209    -6.00    26.0
#  --------------------------------------------------------
#
#
#  Post Hoc Tests - alcohol
#  -------------------------------------------------------------
#                           psi-hat   p         Lower    Upper
#  -------------------------------------------------------------
#    None       2 Pints     -3.33      0.611    -20.5     13.8
#    None       4 Pints     35.83     < .001     19.3     52.3
#    2 Pints    4 Pints     39.17     < .001     22.5     55.9
#  -------------------------------------------------------------
#

---

Robust Descriptives

Description

Robust Descriptives

Usage

rdesc(
  data,
  vars,
  splitBy = NULL,
  mean = TRUE,
  trim = TRUE,
  tr = 0.2,
  win = FALSE,
  wl = 0.2,
  mest = FALSE,
  bend = 1.28,
  med = FALSE
)

Arguments

data	the data as a data frame
vars	a vector of strings naming the variables in data of interest
splitBy	a string naming the variable in data to split the data by
mean	TRUE (default) or FALSE, provide a 'normal' arithmetic mean
trim	TRUE (default) or FALSE, provide a trimmed mean
tr	a number between 0 and 0.5 (default: 0.2); the proportion of measurements to trim from each end when producing trimmed means
win	TRUE or FALSE (default), provide a 'Winsorized' mean
wl	a number between 0 and 0.5 (default: 0.2); the level of 'winsorizing' when producing winsorized means
mest	TRUE or FALSE (default), provide an 'M-estimated' value
bend	a number (default: 1.28), the bending constant to use when using M-estimators
med	TRUE or FALSE (default), provide medians

Value

A results object containing:

results$table					the table of descriptives

Tables can be converted to data frames with asDF or as.data.frame. For example:

results$table$asDF

as.data.frame(results$table)

Examples

data('eurosoccer', package='WRS2')

SpainGermany <- subset(eurosoccer, eurosoccer$League == 'Spain' | eurosoccer$League == 'Germany')
SpainGermany <- droplevels(SpainGermany)

walrus::rdesc(
    data = SpainGermany,
    vars = "GoalsGame",
    splitBy = "League",
    med = TRUE)

#
#  ROBUST DESCRIPTIVES
#
#  Robust Descriptives
#  ----------------------------------------------------------
#                                                    SE
#  ----------------------------------------------------------
#    GoalsGame    Germany    Mean            1.46     0.105
#                            Trimmed mean    1.45    0.1341
#                            Median          1.43    0.1599
#
#                 Spain      Mean            1.45     0.101
#                            Trimmed mean    1.33    0.0601
#                            Median          1.30    0.0766
#  ----------------------------------------------------------
#

---

Box & Violin Plots

Description

Box & Violin Plots

Usage

rplots(
  data,
  vars,
  splitBy = NULL,
  violin = TRUE,
  boxplot = FALSE,
  dot = TRUE,
  dotType = "stack"
)

Arguments

data	the data as a data frame
vars	a vector of strings naming the variables in data of interest
splitBy	a string naming the variable in data to split the data by
violin	TRUE (default) or FALSE, provide violin plots
boxplot	TRUE or FALSE (default), provide box plots
dot	TRUE (default) or FALSE, plot each measurement as a dot
dotType	'jitter' or 'stack' (default); whether data dots are jittered or stacked

Value

A results object containing:

results$plots					an array of images

Examples

data('eurosoccer', package='WRS2')

# violin plots

walrus::rplots(
    data = eurosoccer,
    vars = "GoalsGame",
    splitBy = "League")


# box plots

walrus::rplots(
    data = eurosoccer,
    vars = "GoalsGame",
    splitBy = "League",
    violin = FALSE,
    boxplot = TRUE,
    dot = FALSE)

---

Robust Independent Samples T-Test

Description

Robust Independent Samples T-Test

Usage

rttestIS(
  data,
  deps,
  group,
  yuen = TRUE,
  tr = 0.2,
  mest = FALSE,
  method = "mom",
  yuenbt = FALSE,
  nboot = 599,
  md = FALSE,
  ci = FALSE,
  es = FALSE,
  esci = FALSE
)

Arguments

data	the data as a data frame
deps	a vector of strings naming the dependent variables in data
group	a string naming the grouping variable in data; must have 2 levels
yuen	TRUE (default) or FALSE, use the Yuen's trim method
tr	a number between 0 and 0.5, (default: 0.2), the proportion of measurements to trim from each end, when using the trim and bootstrap methods
mest	TRUE or FALSE (default), use an M-estimator
method	'onestep', 'mom' (default) or 'median', the M-estimator to use; One-step, Modified one-step or Median respectively
yuenbt	TRUE or FALSE (default), use the Yuen's bootstrap method
nboot	a number (default: 599) specifying the number of bootstrap samples to use when using the bootstrap method
md	TRUE or FALSE (default), provide the mean difference
ci	TRUE or FALSE (default), provide a 95% confidence interval on the mean difference
es	TRUE or FALSE (default), provide the effect-size
esci	TRUE or FALSE (default), provide a 95% confidence interval on the effect-size

Value

A results object containing:

results$ttest					the table of t-test results

Tables can be converted to data frames with asDF or as.data.frame. For example:

results$ttest$asDF

as.data.frame(results$ttest)

Examples

data('eurosoccer', package='WRS2')

SpainGermany <- subset(eurosoccer, eurosoccer$League == 'Spain' | eurosoccer$League == 'Germany')
SpainGermany <- droplevels(SpainGermany)

rttestIS(SpainGermany,
         dep = 'GoalsScored',
         group = 'League',
         yuen = TRUE,
         mest = TRUE)

#
#  ROBUST INDEPENDENT SAMPLES T-TEST
#
#  Robust Independent Samples T-Test
#  ---------------------------------------------------------
#                                  t         df      p
#  ---------------------------------------------------------
#    GoalsScored    Yuen's test     0.297    17.3    0.770
#                   M-estimator    -0.933            0.993
#  ---------------------------------------------------------
#

---

Robust Paired Samples T-Test

Description

Robust Paired Samples T-Test

Usage

rttestPS(data, pairs, tr = 0.2, md = FALSE, es = FALSE, ci = FALSE)

Arguments

data	the data as a data frame
pairs	a list of lists specifying the pairs of measurement in data
tr	a number between 0 and 0.5, (default: 0.2), the proportion of measurements to trim from each end, when using the trim and bootstrap methods
md	TRUE or FALSE (default), provide means and standard errors
es	TRUE or FALSE (default), provide effect sizes
ci	TRUE or FALSE (default), provide confidence intervals

Value

A results object containing:

results$ttest					the table of t-test results

Tables can be converted to data frames with asDF or as.data.frame. For example:

results$ttest$asDF

as.data.frame(results$ttest)

Examples

data(anorexia, package='MASS')
anorexiaFT <- subset(anorexia, subset = Treat == "FT")

rttestPS(anorexiaFT,
         pairs = list(
             list(i1 = 'Prewt', i2 = 'Postwt')))

#
#  ROBUST PAIRED SAMPLES T-TEST
#
#  Robust Paired Samples T-Test
#  ---------------------------------------------
#                       t        df      p
#  ---------------------------------------------
#    Prewt    Postwt    -3.83    10.0    0.003
#  ---------------------------------------------
#

---