Simulations investigating the
decision-boundary strategy analyses conducted in the COVIS
literature as reported in
Edmunds, Milton, Wills (accepted).
This DAU contains simulation files, functions, simulation output
and scripts to construct the tables, all in open cross-platform
formats (see the
file format notes).
Issues
2019-02-20: (open). Simulation 1, UD strategy, at
zero noise, sometimes returns an incorrect answer if run on more than 8
cores (the published simulations used 4-8 cores).
Resources
R scripts for simulations and presentation of PLY34. You
will need R, plus
the following packages that are available on
CRAN:
ez, grt, mvtnorm, foreach, doParallel, parallel, mnormt,
ggplot2
PLY34functions.R -
functions required by the other scripts, below.
PLY34Results.R -
Calculate results reported in the Tables and Figure of
the paper, using stored results of the simulations.
Simulation1stored.csv
- The output from Simulation 1. Column headings
are as follows:
categoryStructure: Category structure simulation is
conducted on. II=Information-integration
strategyType: The strategy type of the simulated responses.
UD=Unidimensional, CJ=Conjunction,
GLC=Diagonal (General linear classifier)
perceptualNoise: The amount of perceptual noise
added to the participants' responses
boundaryNoise: The amount of boundary noise added
to the participants' responses
no_ppts: The number of stimulated participants that
were averaged on that row
prop_UD: The proportion of participants that were
found to be best described by a unidimensional
strategy
prop_GLC: The propotion of participants that were
found to be best described by a diagonal (GLC)
strategy
prop_CJ: The proportion of participants that were
found to be best described by a conjunction
strategy
prop_RND: The proportion of participants that were
found to be best described by a random strategy
wBIC_UD: The averaged wBIC for participants who
were best described by a unidimensional strategy
wBIC_GLC: The averaged wBIC for participants who
were best described by a diagonal (GLC) strategy
wBIC_CJ: The averaged wBIC for participants who
were best described by a conjucntion strategy
wBIC_RND: The averaged wBIC for participnats who
were best described by a random strategy
Simulation2stored.csv -
The output from Simulation 2. The column headings are
identical to those in Simulation 1 above. The only
change is that categoryStructure is equal to
UD=Unidimensional
Simulation3stored.csv -
The output from Simulation 3. Each row is a single
simulated participant. Column headings are as
follows:
categoryStructure: Category structure simulation
is conducted on. II=Information-integration,
UD=unidimensional
strategyType: The strategy type of the simulated
responses. UD=Unidimensional, CJ=Conjunction,
GLC=Diagonal(General linear classifier)
perceptualNoise: The amount of perceptual noise
added to the participant's responses
boundaryNoise: The amount of boundary noise added
to the participant's responses
Accuracy: The proportion accuracy of that
participant
UDX: Whether the participant was recovered as
using a unidimensional strategy on the x-axis: 0=no,
1=yes
UDY: Whether the participant was recovered as
using a unidimensional strategy on the y-axis: 0=no,
1=yes
GLC: Whether the participant was recovered as
using a diagonal (GLC) strategy: 0=no, 1=yes
CJ: Whether the participant was recovered as
using a conjunction strategy: 0=no, 1=yes
RND: Whether the participant was recovered as
using a random strategy: 0=no, 1=yes
wBIC_UDX: The wBIC for the unidimensional
strategy on the x-axis
wBIC_UDY: The wBIC for the unidimensional
strategy on the y-axis
wBIC_GLC: The wBIC for the diagonal (GLC)
strategy