The files ARL2-bn-Ss.dat and ARL2-wn-Ss.dat contain the "average response levels (ARLs)"
for the individual subjects from Experiment 1 reported in the following publication:

Petrov, A.  (2010). Category rating is based on prototypes and not instances: 
  Evidence from feedback-dependent context effects. 
  Journal of Experimental Psychology: Human Perception and Performance, ?, ??-??.   
PDP preprint available at http://alexpetrov.com/pub/jephpp10/

This experiment was developed and stored under the name "fdbk1".
See fdbk1/data/raw_data/ for the raw data.  
The raw data were preprocessed using fdbk1/data/data_by_sbj.m and stored in a
Matlab structure S, which in turn was stored in file fdbk1/data/S.mat.

The Matlab script fdbk1/data/empir_ARL2_profile.m loads S.mat, augments
it with a new field 'ARL2' and stores it in file fdbk1/data/S2.mat.  The new
field aggregates the ARLs in 56-trial windows (the original ARL calculated by
fdbk1/data/data_by_sbj.m are based on 28-trial windows).  All figures and analyses
in the published paper are based on the 56-trial average response levels, which
are stored in S2(sbj).ARL2.  All the other data is available in the S2 structure too.
In general, S2 contains all the information available in the raw ASCII files, plus
the ARLs calculated on this basis. All subsequent scripts load S2.mat as their starting point.

S2 is a Matlab "structure array" with one element per subject and the following fields.
(The record for the first subject is printed as an example.)
>> S2(1)
ans = 
    sbj_number: 1
         group: 3
      miss_idx: []
          dist: [476x1 double]    % stimulus length -- the distance b/n the dots [in pixels]
         categ: [476x1 double]    % category -- from 1 to 7
          fdbk: [476x1 double]    % 0="X", 1-7
          resp: [476x1 double]    % the key pressed by the subject
            RT: [476x1 double]    % response time in milliseconds
        interv: [476x1 double]    % auxiliary, used in ARL calculation
           ARL: [17x1 double]     % average response levels, calculated in 28-trial blocks
          ARL2: [9x1 double]      % ARL averaged two by two. See empir_ARL2_profile.m


The script fdbk1/data/empir_ARL2_profile.m was published to html using Matlab's
"publish" functionality.  The published report includes the source code, all messages
printed to the command window, and all figures. Figures 2 and 3 in the JEP:HPP article were
generated by this script -- the html is available in fdbk1/data/html/empir_ARL2_profile.html 

The ARL2s for each subject in each of the 9 time periods were exported in ASCII format in
order to be imported into SPSS for ANOVA.  This was again accomplished by empir_ARL2_profile.m.
The results were stored in fdbk1/data/ARL2-bn-Ss.dat and ARL2-wn-Ss.dat as follows:

fdbk1/data/ARL2-bn-Ss.dat has one row per data point and the following columns:
    * 1. Sbj -- 1-55
    * 2. Group -- 1=U1, 2=L1, 3=H1, 5=L2, 6=H2
    * 3. Context -- 1=U, 2=L, 3=H
    * 4. Order -- 1=Feedback_first, 2=No_feedback_first
    * 5. Block -- 1-9. Each is 56 trials (except the first, which is 28).
    * 6. Feedback -- 0=None, 1=Fdbk
    * 7. ARL2

fdbk1/data/ARL2-wn-Ss.dat rearranges exactly the same data in within-subject format
expected by SPSS repeated measures analysis.  It has one row per subject and the following columns:
    * 1. Sbj -- 1-55
    * 2. Group -- 1=U1, 2=L1, 3=H1, 5=L2, 6=H2
    * 3. Context -- 1=U, 2=L, 3=H
    * 4. Order -- 1=Feedback_first, 2=No_feedback_first
    * 5. W -- ARL2 value for the "warm-up" initial block
    * 6-9. F1-F4 -- ARL2 values during the four feedback blocks
    * 10-13. N1-N4 -- ARL2 values during the four no-feedback blocks

Note that the within-subject data format throws away the temporal order of observations. They are rearranged depending on Order so that in the resulting ASCII file all feedback ARLs are listed in columns F1-F4 and all no-feedback ARLs are listed in columns N1-N4.

Results of the within-subject ANOVA are available in fdbk1/data/SPSS/

-- Alex Petrov  2010-07-26
