Simulation

• A Simulation of Regression with Random Values
  This is SPSS code from Terry.Taerum@ualberta.ca to create simulated data for 40 variables, It loops to randomly assign a 1 to var40 for the first 10 cases and a 0 for the remainder. This part of the program can be used to create randomly initialized variables for any subsequent statistical procedure. In this case, the simulation then executes a regression and a logistic regression for the created variables. This simulation demonstrates that if you use 39 independents in a regression, one or two will be found to be significant in most runs of the simulation, even though the data are, in fact, random.

  new file.
  input program.
  numeric v1 to v40.
  vector vx=v1 to v39.
  loop #1=1 to 648.
  if (#1 le 10) v40=1.
  if  (#1 gt 10) v40=1.
  loop #j=1 to 39.
  compute vx(#j)=trunc(5*rv.uniform(0,1)).
  end loop.
  end case.
  end loop.
  end file.
  end input program.
  execute.
  regression vars=v1 to v40
    /dependent=v40/enter v1 to v39.
  logistic regression var=v40
     /method=enter v1 to v39.
  
  

Bibliography and Links

Links

• The Simulation Book, Bill Troachim, Cornell University

  Bibliography
  

• Pool, Ithiel de Sola, R. Abelson, and S. Popkin (1965). Candidates, issues, and strategies: A computer simulation of the 1960 and 1964 presidential elections. Cambridge, MA: MIT Press. Classic work used in the Kennedy campaign in 1960. Pool developed a model of voting which predicted election results with greater accuracy than previously possible.

• Shaffer, William R. (1962). Computer simulations of voting behavior. NY: Oxford. Another classic.