Examples of SOS in Action


Below are descriptions of each example script reported in the paper.  For each example, a .zip file may be downloaded that contains the script file ("<example>.m") and the population file(s) used by the script (e.g., "MRC.txt"). We have also included a file containing the success rate of each example from 10 independent runs of the SOS software ("success.txt"); the final samples from each successful run are contained within the "results" folder.

Full sets of examples are available as single archives following the individual examples.


'ME95' and Entropy Analogue Designs from the SOS article

Users may want to start by studying these scripts and the details of these optimizations, as described in the main SOS manuscript.


One-way designs

  • One-way, 2 samples, groupwise: this script creates 2 samples of 100 words each from the MRC database file.  The 2 samples will be different groupwise on 1 dimension ('KFfrequency'), p-value < .05, and matched groupwise on 3 dimensions ('letters', 'syllables', and 'phonemes'), p-value > 0.5.  The optimization uses the stochastic method of annealing.   [download .zip] 
  • One-way,  2 samples, pairwise: this script creates 2 samples of 100 words each from the MRC database file.  The 2 samples will be different pairwise on 1 dimension ('KFfrequency'), p-value < .05, and matched pairwise on 3 dimensions ('letters', 'syllables', and 'phonemes'), p-value > 0.5.  The optimization uses the stochastic method of annealing. [download .zip]
  • One-way, 3 samples, groupwise: this script creates 3 samples of 100 words each from the MRC database file.  The 3 samples will be different groupwise on 1 dimension (high, medium, and low 'KFfrequency'), p-value < .05, and matched groupwise on 3 dimensions ('letters', 'syllables', and 'phonemes'), p-values > 0.5.  The optimization uses the stochastic method of annealing.  [download .zip]
  • One-way, 3 samples, pairwise: this script creates 3 samples of 100 words each from the MRC database file.  The 3 samples will be different pairwise on 1 dimension (high, medium, and low 'KFfrequency'), p-value < .05, and matched pairwise on 3 dimensions ('letters', 'syllables', and 'phonemes'), p-values > 0.5.  The optimization uses the stochastic method of annealing. [download .zip]

'locked set' designs

  • Locked set, groupwise: this script creates 1 sample of 100 words from a subset of the MRC database that includes only nouns.  The 1 sample is matched groupwise on 4 dimensions ('KFfrequency', 'letters', 'syllables', and 'phonemes') to 100 pre-selected verbs of frequency < 50 from the MRC database.  This optimization uses the stochastic method of annealing.  [download .zip]
  • Locked set, pairwise: this script creates 1 sample of 100 words from a subset of the MRC database that includes only nouns.  The 1 sample is matched pairwise on 4 dimensions ('KFfrequency', 'letters', 'syllables', and 'phonemes') to 100 pre-selected verbs of frequency < 50 from the MRC database.  This optimization uses the stochastic method of annealing.  [download .zip]

'Match to a value' designs

  • Match to a value, groupwise:  this script creates 2 samples of 100 words each from the MRC database file.  For 1 sample, 'KFfrequency' values must be > 100, and for the other sample, 'KFfrequency' values must be < 10.  The 2 samples are maximally different groupwise on 1 dimension ('KFfrequency'), p < .05, and are matched groupwise on 3 dimensions ('letters', 'syllables', and 'phonemes'), p-values > 0.5.  This optimization uses the stochastic method of annealing. [download .zip]
  • Match to a value, pairwise: this script creates 2 samples of 100 words each from the MRC database file.  For 1 sample, 'KFfrequency' values must be > 100, and for the other sample, 'KFfrequency' values must be < 10.  The 2 samples are maximally different pairwise on 1 dimension ('KFfrequency'), p < .05, and are matched pairwise on 3 dimensions ('letters', 'syllables', and 'phonemes'), p-values > 0.5.  This optimization uses the stochastic method of annealing. [download .zip]


Additional 2 x 2 examples:


  • 2x2 design: This example illustrates a 2x2 design where all of the observations are sampled from a single population and SOS discovers the optimal separation point. [download .zip]


Full sets of Examples


All examples in the SOS manuscript

   Download all examples described in the SOS manuscript


Additional Examples

We have also made available many other example scripts, most notably of examples of "entropy" constraints.  To download all additional examples (including their respective population files):
   
   Download all additional examples






Blair Armstrong, Christine Watson, David Plaut, 2011-2012