Vocabulary Networks: Program-15
Testing for attrition

Use these boxes to set the basic parameters for your model

NTWK sets up the network for this simulation. Choose a value between 0000 and 9999 for this parameter.

nEv sets the number of attrition events in the simulation. Choose a value between 0 and 500 for this parameter.

  rEv determines which words are affected by your instructions. Choose a value between 0 and 9999 for this parameter.


  mEv spread out the attrtition events. Choose a value between nEv and 400 for this parameter.


  Twds A list of 20 numbers between 0 and 999. The program will test these words at each test event


  tEv a list of test times e.g. choose values between 100 and 500 for this parameter

This simulation explores how accurately a small test
can measure attrition in a vocabulary network.
The program builds a 1000 word network and lets it
settle into a stable attactor state. The program then
runs a series of attrition events distributed over
mEv updates. The program assess how much attrition has
occurred by testing the activation status of a pre-
determined set of test words.
The parameters let you vary the tested words and when
testing events take place.

THINGS TO DO.
Set nEv to 100, and set mEv to 400.
Enter a list of 20 numbers to parameter Twds.
Set tEv to "200,300,400".
Choose random values for NTWK and rEv.
How does the network respond to these parameter values?
How accurately does the test capture attrition events?
What efffects emerge when you vary the parameter values?
What happens when the network has few active words?
Would you get the same results with a larger network?

QUESTIONS TO ASK.
What features does the test format capture well?
What features does the test format miss?
What is the optimum number of test events?
How can you optimise a small number of test events?
How does the number of active words in the network
affect the performance of the test?