PDA

View Full Version : False positives


ThinkQuick
07-25-2005, 05:56 PM
There was a maxim article that said that polygraphs (lie detectors) will catch you lying 83% of the time, but will also say that an honest person is lying 40% of the time.
I don't really care whether this is true, just wondering how you would calculate the probability someone is lying based on these facts and a 'LIAR' reading on the polygraph.

BruceZ
07-25-2005, 06:04 PM
[ QUOTE ]
There was a maxim article that said that polygraphs (lie detectors) will catch you lying 83% of the time, but will also say that an honest person is lying 40% of the time.
I don't really care whether this is true, just wondering how you would calculate the probability someone is lying based on these facts and a 'LIAR' reading on the polygraph.

[/ QUOTE ]

That depends on what percentage of people lie.

bobman0330
07-25-2005, 06:57 PM
Agreed. The probability = (times the polygraph goes off and subject is lying)/(times polygraph goes off total), with the denominator = true positives + false positives = .83 x + .4 (1-x), where x is the probability that a person is lying. So, .83 x / (.83x + .4(1-x))

BettyBoopAA
07-25-2005, 07:51 PM
.83/(.83+.4)= 67% chance of lying

PairTheBoard
07-25-2005, 08:52 PM
[ QUOTE ]
.83/(.83+.4)= 67% chance of lying

[/ QUOTE ]

That's if the polygrapher gives him a 50-50 chance before the test. But since the polygrapher KNOWS the guy is guilty as sin he calculates it:

.83(1)/[.83(1)+.4(1-1)] = 100%

PairTheBoard

SheetWise
07-25-2005, 10:25 PM
I now think I understand bias /images/graemlins/wink.gif