| Basic Commands | Probability Puzzles | Hypothesis Test, Count Data | Hypothesis Test, Measured Data | Confidence Interval, Count Data | Confidence Interval, Measured Data | Association / Correlation | Regression | Other Examples |

Cancer Medicine Effectiveness

A new medicine produced 5 cures in 6 patients, while a placebo produced cures in only 2 in 6. If the medicine is ineffective, what is the probability that this difference might occur by chance. We will generate two samples (6 random numbers between 1 and 12) and count the number of cures in each (our benchmark or null hypothesis is that both the placebo and medicine produce an overall cure rate of 7 in 12, so we let the numbers 1-7 represent cures).

URN 7#1 5#0 a

constitute the null universe

REPEAT 1000
  SAMPLE 6 a b

take two samples like those observed, from the null model

  SAMPLE 6 a c
  COUNT b =1 bb

count the number of cures

  COUNT c =1 cc
  SUBTRACT bb cc d

what's the difference in # of cures?

 SCORE d z

keep score

END
HISTOGRAM z            
COUNT z >=3 k1

how often a difference of 3 cures in either direction?

COUNT z <=-3 k2
ADD k1 k2 k
DIVIDE k 1000 kk
PRINT KK