NCA (version 4.0.1)

nca_analysis: Run NCA analyses on a data set

Description

Run multiple types of NCA analyses on a dataset

Usage

nca_analysis(data, x, y, ceilings=c('ols', 'ce_fdh', 'cr_fdh'),
             corner=NULL, flip.x=FALSE, flip.y=FALSE, scope=NULL,
             bottleneck.x='percentage.range', bottleneck.y='percentage.range',
             steps=10, step.size=NULL, cutoff=0, qr.tau=0.95,
             effect_aggregation = 1, test.rep=0,
             test.p_confidence=0.95, test.p_threshold=0.05)

Value

Returns a list of 6 items (see examples for further explanation):

plots

A list of plot-data for each x-y combination

summaries

A list of dataframes with the summaries for each x-y combination

bottlenecks

A list of dataframes with a bottleneck table for each ceiling technique

peers

A list of ceilings, with a list of peers for each independent variable. Peers are corner points of the CE-FDH ceiling line (e.g., the northwest-corners points for corner = 1)

tests

The results of the test for each independent variable (not human friendly, use nca_output)

test.time

The total time needed to run the tests for all independent variables

Arguments

data

dataframe with columns of the variables

x

index or name (or a vector of those) with independent variable(s) x

y

index or name of the column with the dependent variable y

ceilings

vector with the ceiling techniques to include in this analysis

corner

either an integer or a vector of integers, indicating the corner to analyze, see Details

flip.x

reverse the direction of the independent variables
Use either a boolean for all independent variables, or a vector with the same length as x

flip.y

reverse the direction of the dependent variables, boolean

scope

a theoretical scope in list format : (x.low, x.high, y.low, y.high), see Details

bottleneck.x

options for displaying the independent variables in the bottleneck table
'percentage.range' to display the percentage of range between min(x) and max(x)
'percentage.max' to display the percentage of max(x)
'actual' to display the actual values
'percentile' to display the percentiles
Using percentage.max with negative values might yield counterintuitive results.

bottleneck.y

options for displaying the dependent variables in the bottleneck table.
See bottleneck.x

steps

this argument accepts 2 types :
- an integer (number of steps in the bottleneck table)
- a list of values (used as Y values in the bottleneck table).
Only useful if step.size is not defined.

step.size

define the step size in the bottleneck table.
The user will be warned if the stepsize does not fit the Y range.
Defaults to null for using steps.

cutoff

display calculated x,y values that are lower/higher than lowest/highest observed x,y values in the bottleneck table as:
0 : NN (not necessary) and NA (not available)
1 : NN (not necessary) and highest observed values
2 : calculated values

qr.tau

define the qr tau (between 0 and 1) for the quantile regression ceiling technique, default 0.95

effect_aggregation

define the corners to aggregate into the effect size. 1 is upper-left and is always selected, 2 is upper-right, 3 is lower-left and 4 is lower-right

test.rep

number of resamples in the statistical approximate permutation test. For test.rep = 0 no statistical test is performed

test.p_confidence

confidence level of the estimated p-value.
Is used to calculate the p-accuracy for a given number of resamples (test.rep), default 0.95.

test.p_threshold

define the threshold significance level in the returned plot of the statistical test, default 0.05

Details

Corners
Corner 1 is the upper-left corner and corner 2 is the upper-right corner. These two corners are used for an analysis of the necessity of the presence/high level if x (corner = 1 ) or the absence/low level if x (corner = 2) for the presence/high level of y, respectively.
Corner 3 is the lower-left corner and corner 4 is the lower-right corner. These two corners are used for an analysis of the necessity of the presence/high level of x (corner = 3 ) or the absence/low level if x (corner = 4) for the absence/low level of y, respectively.
By default the upper left corner is analysed for all independent variables and corner is not defined. If corner is defined, flip.x and flip.y are ignored.

Scope
By default, the theoretical scope is not defined and the empirical scope is used based on the minimum and maximum observed values of x and y.

Examples

# Load the data
data(nca.example)
data <- nca.example

# Basic NCA analysis, with independent variables in the first 2 columns
# and the dependent variable in the third column
model <- nca_analysis(data, c(1, 2), 3)

# Use nca_output to show the summaries (see nca_output documentation for more options)
nca_output(model)

# Columns can be selected by name as well
model <- nca_analysis(data, c('Individualism', 'Risk taking'), 'Innovation performance')

# Define the ceiling techniques via the ceilings parameter, see 'ceilings' for all types
model <- nca_analysis(data, c(1, 2), 3, ceilings=c('ce_fdh', 'ce_vrs'))

# These are the available ceiling techniques
print(ceilings)

# By default the upper-left corner is analysed. With the corner argument for each
# independent variable a different corner can be selected. Select corner 1 or 2
# for an analysis of necessary conditions for the presence/high level of the
# dependent variable, and corner 3 or 4 for an analysis of necessary conditions for
# the absence/low level of the dependent variable. It is not possible to combine
# corner 1 or 2 with corner 3 or 4 in the same analysis as different outcomes are analysed.
# This analyses the upper right corner for the first independent variable
# and the upper left corner for the second independent variable:
model <- nca_analysis(data, c(1, 2), 3, corner=c(2, 1))

# Alternatively, for using the upper right corner(s), 'flip' the x variables
model <- nca_analysis(data, c(1, 2), 3, flip.x=TRUE)

# It is also possible to flip a single x variable
model <- nca_analysis(data, c(1, 2), 3, flip.x=c(TRUE, FALSE))

# Flip the y variable if the lower corners need analysing
model <- nca_analysis(data, c(1, 2), 3, flip.x=c(TRUE, FALSE), flip.y=TRUE)

# Use a theoretical scope instead of the (calculated) empirical scope
model <- nca_analysis(data, c(1, 2), 3, scope=c(0, 120, 0, 240))

# Display the peers for a ceiling and an independent variable
print(model$peers$ce_fdh$Individualism)

# By default, the bottleneck tables use percentages of the range for the x and y values.
# Using the percentage of the max value is also possible
model <- nca_analysis(data, c(1, 2), 3, bottleneck.y='percentage.max')

# Use the actual values, in this case the x-value
model <- nca_analysis(data, c(1, 2), 3, bottleneck.x='actual')

# Use percentile, in this case for the y-values
model <- nca_analysis(data, c(1, 2), 3, bottleneck.y='percentile')

# Any combination is possible
model <- nca_analysis(data, c(1, 2), 3, bottleneck.x='actual', bottleneck.y='percentile')

# The number of steps is adjustible via the steps parameter
model <- nca_analysis(data, c(1, 2), 3, steps=20)

# The steps parameter also accepts a list of values
# These are interpreted as actual or percentage / percentile depending on bottleneck.y
model <- nca_analysis(data, c(1, 2), 3, steps=seq(50, 120, 10))

# Or via the step.size parameter, this ignores the steps parameter
model <- nca_analysis(data, c(1, 2), 3, step.size=5)

# If the ceiling line crosses the X = Xmax line at a point C below Y = Ymax,
# for Y < Yc < Ymax, the corresponding X in the bottleneck table is displayed as 'NA'
# It is also possible to display them as Xmax
model <- nca_analysis(data, c(1, 2), 3, cutoff=1)

# or as the calculated value on the ceiling line
model <- nca_analysis(data, c(1, 2), 3, cutoff=2)

# To run tests, the test.rep needs to be larger than 0
# Optionally the p_confidence (default 0.95) and the p_threshold (default 0) can be set
model <- nca_analysis(data, c(1), 3, test.rep=1000, test.p_confidence=0.9, test.p_threshold=0.05)
model <- nca_analysis(data, c(1), 3, test.rep=100, test.p_confidence=0.9, test.p_threshold=0.05)
# The output of the tests can be shown via nca_output with test=TRUE
nca_output(model, test=TRUE)