This code includes an interactive data browser that lets you explore a curated version of the dataset. However, because Matlab has made changes to how figures and callback functions are handled in recent versions, the data browser is very dodgy for Matlab versions earlier than 2014b. If you have an older version of Matlab, please see the "The data" section for details about accessing the raw data itself.
In order to use the browser, clone/download the repository and run
PlotCurated
which will open an interactive figure with two subplots. The left subplot shows data for 5% density, while the right subplot shows for 24% density. Each point in the plot shows a single cell. By default, the axes will show the r-value (correlation coefficient) between the anticorrelated and correlated tuning curves on the y-axis, and between the half-matched and correlated tuning curves on the x-axis. You can change the axes by selecting "Plot -> Main: x axis", and choosing an appropriate metric (for example half-matched slope). Most metrics should be self-explanatory upon reading Henriksen, Read, & Cumming (2016), but should you have any questions about the data, please get in touch [sid(dot)henriksen(at)gmail(dot)com].
In the plots, each unique colour shows a different recording session. The size of the points shows the magnitude of the Disparity Discrimination Index. Square points correspond to cells from monkey Jbe, while circular points correspond to cells from monkey Lem. Clicking the dots will bring up tuning curves showing that cells' responses to correlated (red line), anticorrelated (black line) and half-matched stereograms (blue line). The disparity tuning curve plot also allows you to toggle different error bars (SEM, SD, and 95% bootstrap CIs).
The data browser requires Matlab version 2014b or later. If you don't have access to this (or you don't have Matlab at all),
you can access the curated data itself and explore it manually. The data is stored in the file CuratedCells.mat. If you run
load('CuratedCells.mat')
while in the repo directory, a variable called Base will be loaded into your workspace. Base is a struct with fields
density, Cells, exptlist, penetrationlist. Base is a 1x2 array, where the elements correspond to the two densities
used (5% and 24%).
density gives the density of the random dot stereograms used (5% or 24%).
Cells is the struct containing all the curated data.
exptlist lists all the recording sessions
penetrationlist gives the x-y coordinates for each recording session.
This contains all the curated data for these experiments. To access this data:
density5_cell1 = Base(1).Cells(3);
This gives you the data for the 1st cell, where the dot density was 5%. Base(1) corresponds to 5% density recordings, and Base(2) to 24% density recordings. To plot a tuning curve we can do:
current_cell = Base(1).Cells(1);
dx = current_cell.Dx;
correlated = current_cell.correlatedResponse;
anticorrelated = current_cell.anticorrelatedResponse;
halfmatched = current_cell.halfmatchedResponse;
figure();
plot(dx,correlated,'r -',dx,halfmatched,'- b',dx,anticorrelated,'k -','linewidth',3);
xlabel('Disparity (deg)');
ylabel('Spike count');
The following gives a complete documentation of the fields in Base(k).Cells:
cellnumber - the cell number given in the recording session.
filename - where the data is located locally (not useful for external use).
regHm - output of a type 2 regression between correlated and half-matched tuning curves.
This is a vector of size three, where the entries are [r,m,b]. r is the correlation coefficient,
m is the half-matched slope, b is the offset (probably useless).
regAc - same as regHm just for anticorrelated slope.
regHmRegular - same as regHm except using OLS regression.
regAcRegular same as regAc except using OLS regression.
Dx - disparities used in the experiment.
correlatedResponse - tuning curve for correlated stimuli (averaged across trial)
halfmatchedResponse - tuning curve for half-matched stimuli
anticorrelatedResponse - tuning curve for anticorrelated stimuli
correlatedSEM - SEMs for the correlated tuning curve
halfmatchedSEM - SEMs for the half-matched tuning curve
anticorrelatedSEM - SEMs for the anticorrelated tuning curve
RMS - root mean square for tuning curves for correlated, half-matched, and anticorrelated stimuli
DDI - DDI computed on correlated tuning curves
HMauc - area under the ROC curve (AUROC) for half-matched stimuli
HMdprime - HMauc converted to a d' value
Cauc - AUROC for correlated stimuli
Cdprime - Cauc converted to a d' value
HMaucZ - AUROC for half-matched stimuli (spike counts Z-normalised by block number to correct for slow drifts)
CaucZ - same as HMaucZ for correlated stimuli
density - density of the stimulus
dw - dot width used (in degrees)
ciLowHm and ciHighHm - 95% bootstrap confidence intervals (CIs) for half-matched tuning curves
ciLowC and ciHighC - 95% bootstrap CIs for correlated tuning curves
ciLowAc and ciHighAc 95% bootstrap CIs for anticorrelated tuning curves
CHm_r_CI - 95% bootstrap CIs for r between correlated and half-matched
CHm_slope_CI - 95% bootstrap CIs for half-matched slope
CAc_r_CI - 95% bootstrap CIs for r between correlated and anticorrelated
CAc_slope_CI - 95% bootstrap CIs for anticorrelated slope
DDIhm - half-matched DDI