Testing Your Program

“The productive modeler simply assumes that software mistakes are inevitable and continually searches for them.”
- Railsback & Grimm

• Understand the difference between validation and verification.
• Seven common kinds of software errors that will blow your mind!
• Ten important tecnhiques for finding and fixing software errors everyone should know.
• Understand why and how to document software tests.

• Typographical Errors
• Syntax Errors
• Misunderstanding Primitives: Show example
• Wrong Display Settings: Use resize-world in setup!
• Logic Errors: Runs but results incorrect
• Run-time Errors: No syntax or logic errors, but breaks on Go (sometimes)
• Formulation Errors: Incorrect assumptions & model decisions

Syntax Checking: Chunk-it and use skeleton code!

ifelse (xcor >= min-marriage-age)
[ show "If" ]
[ show "Else" ]

ifelse (xcor >= min-marriage-age) and
       (random-float 1.0 < 0.1)
[ show "If" ]
[ show "Else" ]

ifelse (xcor >= min-marriage-age) and
       (random-float 1.0 < 0.1)
[ set married? true ]
[ show "Else" ]

Visual Testing: Use visual cues of variables!

• Use scale-color to color turtles or patches based on their variables.
• Use label and plabel to check turtle or patch information.
• Use Agent and Patch Monitor.
• Use a smaller World to test things (actually, testing on smaller problem is a more general technique)
• Slow down the simulation and/or use a step button.

Print Statements

• For procedures:

to dostuff
  show "Starting procedure X"
  ; Do stuff
  show "Ending procedure X"
end

• For variables:

observer> show word "num turtles = " count turtles
observer: "num turtles = 0"

Code Reviews

• Reviewer's job:
  • Verification: Does the code match the ODD model formulation?
  • Fresh set of eyes can more easily find bugs (sometimes).
  • Make sure code is well-organized and easy to understand.
  • Write code as if someone will eventually read AND USE IT, even if you do not plan on it being used.

Statistical Analysis of File Output

Edit: This slide has been modified from lecture to correct an error.

Question: Is the probability butterflies move to the highest neighbor patch really \( q \)?

Answer: No. It is the approximate proportion \[ q + \frac{1-q}{8}. \]

For \( q = 0.4 \), we would expect the butterfly to move to the highest neighbor patch with probability 0.475.

Statistical Analysis of File Output

file-type
file-print
file-open
file-close

Statistical Analysis of File Output

mydata <- read.csv("TestOutput.csv", header=FALSE)
str(mydata)

'data.frame':   1000 obs. of  9 variables:
 $ V1: num  15.8 14.9 17.1 17.8 16.4 ...
 $ V2: num  15.6 15.6 15.6 17 16.3 ...
 $ V3: num  16.3 15.5 16.9 18.5 17.1 ...
 $ V4: num  14.7 17 18.4 19.2 18.4 ...
 $ V5: num  15.6 15.7 16.4 19.1 16.9 ...
 $ V6: num  14.8 16.3 17.7 19.8 17.7 ...
 $ V7: num  15.5 16.4 16.3 17.7 15.6 ...
 $ V8: num  15.5 15.5 17.8 18.3 17.8 ...
 $ V9: num  15.6 17 18.4 17 15.6 ...

Statistical Analysis of File Output

Note: There were errors here during lecture. The code below has been changed to correct the error. I will discuss in class on Wednesday.

moved.to.highest <- sapply(1:1000, function (x) {max(mydata[x,1:8]) == mydata[x,9]})

moved.to.highest <- as.integer(moved.to.highest)

Statistical Analysis of File Output

prop.test(sum(moved.to.highest), 1000)

    1-sample proportions test with continuity correction

data:  sum(moved.to.highest) out of 1000, null probability 0.5
X-squared = 23.409, df = 1, p-value = 1.31e-06
alternative hypothesis: true p is not equal to 0.5
95 percent confidence interval:
 0.3922385 0.4543604
sample estimates:
    p 
0.423