Sunday, January 24, 2010
Bad, bad science. Just...bad
From www.sexhelp.com, linked to by a recent Guardian article...
"SEXUAL ADDICTION SCREENING TEST
We have compared your answers with people who have been diagnosed with sex addiction. Your answers HAVE MET a score on basis of six the criteria that indicate sex addiction is present. To help you understand, the graphic below plots your score in relation to the scores of others.
In addition there are certain subscales to further confirm that a problem exists. The following patterns emerged in your answers:
· A profile consistent with men who struggle with sexually compulsive behavior
· A profile consistent with women who struggle with sexually compulsive behavior
The SAST measures key characteristics of addiction. The following dimensions of an addictive disorder appeared in your answers:
· Preoccupation: obsessive thinking about sexual behavior, opportunities, and fantasies
· Relationship disturbance: sexual behavior has created significant relationship problems
· Affect disturbance: significant depression, despair, or anxiety over sexual behavior
WHAT SHOULD I DO NOW?
You've taken the test and it confirmed your fears. You're probably frightened, confused, and overwhelmed. Where do you go? Whom can you trust?...."
The graph won't reproduce here, so I shall describe it. X axis is "SAST core score" from 0 to 20. Y axis is unlabelled, 0 to 0.20.
There are two curves plotted; non addicts, which starts at around (0,0.18) and goes in a wonky fashion down to hit the x axis at around (11,0) . Addicts, which starts at (0,0) and increases smoothly to reach (11, 0.05), then rises steeply to peak at (15, 0.15) and falls sharply down to (19,0).
These curves cross at around (7.5, 0.05).
I have a problem with the Y axis. The two curves both peak at around 0.15 and cover roughly the same area under each curve. Which suggests at first sight that there must be roughly equal numbers of addict and non-addict participants in the data set. But according to the webpage, estimates of numbers of addicts range from 3% to 6% of the population. So is the data set seriously skewed, or is this being plotted on two different scales?
The areas under each curve are going to approximate to roughly 1 (xthingtimesything) so I am going to conclude that the Y axis reflects the percentage of each population that record each score.
There are some tentative conclusions that one can draw from this graph, without needing to even consider whether the actual diagnoses behind the data are crap.
Firstly, the scores for non addicts fall in the range between 0 and 11. The scores for addicts fall in the range between 0 and 19. It is not therefore possible to draw any firm conclusion from a score between 0 and 11.
Secondly, the score of 6, which generates the rather exciting message above, comes out on the 0.025 point for the addicts scale and 0.08 for the non addicts scale. Which I thought at first meant that a score of 6 is three times more likely to belong to a non addict than an addict.
But it's better than that, of course. Because we are presumably dealing with very different size data sets. Assuming, generously, that there are 90 non addicts in the data for every 10 addicts (a 10% addiction rate). If there were 1000 participants, then there would be 900 non addicts, and 8% of them would score 6. Say 72 of them. Of the 100 addicts, 2.5% would score 6. Say 3.
So a random person comes along and takes this test, and scores 6. Does the website helpfully explain that while the results cannot possibly be conclusive there is only a 4% chance that someone scoring 6 is a sex addict?
Oddly enough, No.
You've taken the test and it confirmed your fears. You're probably frightened, confused, and overwhelmed. Where do you go? Whom can you trust?...."
Not a clue. But I know who you can't. Hello, Noted psychologist and author Dr Patrick Carnes.
"SEXUAL ADDICTION SCREENING TEST
We have compared your answers with people who have been diagnosed with sex addiction. Your answers HAVE MET a score on basis of six the criteria that indicate sex addiction is present. To help you understand, the graphic below plots your score in relation to the scores of others.
In addition there are certain subscales to further confirm that a problem exists. The following patterns emerged in your answers:
· A profile consistent with men who struggle with sexually compulsive behavior
· A profile consistent with women who struggle with sexually compulsive behavior
The SAST measures key characteristics of addiction. The following dimensions of an addictive disorder appeared in your answers:
· Preoccupation: obsessive thinking about sexual behavior, opportunities, and fantasies
· Relationship disturbance: sexual behavior has created significant relationship problems
· Affect disturbance: significant depression, despair, or anxiety over sexual behavior
WHAT SHOULD I DO NOW?
You've taken the test and it confirmed your fears. You're probably frightened, confused, and overwhelmed. Where do you go? Whom can you trust?...."
The graph won't reproduce here, so I shall describe it. X axis is "SAST core score" from 0 to 20. Y axis is unlabelled, 0 to 0.20.
There are two curves plotted; non addicts, which starts at around (0,0.18) and goes in a wonky fashion down to hit the x axis at around (11,0) . Addicts, which starts at (0,0) and increases smoothly to reach (11, 0.05), then rises steeply to peak at (15, 0.15) and falls sharply down to (19,0).
These curves cross at around (7.5, 0.05).
I have a problem with the Y axis. The two curves both peak at around 0.15 and cover roughly the same area under each curve. Which suggests at first sight that there must be roughly equal numbers of addict and non-addict participants in the data set. But according to the webpage, estimates of numbers of addicts range from 3% to 6% of the population. So is the data set seriously skewed, or is this being plotted on two different scales?
The areas under each curve are going to approximate to roughly 1 (xthingtimesything) so I am going to conclude that the Y axis reflects the percentage of each population that record each score.
There are some tentative conclusions that one can draw from this graph, without needing to even consider whether the actual diagnoses behind the data are crap.
Firstly, the scores for non addicts fall in the range between 0 and 11. The scores for addicts fall in the range between 0 and 19. It is not therefore possible to draw any firm conclusion from a score between 0 and 11.
Secondly, the score of 6, which generates the rather exciting message above, comes out on the 0.025 point for the addicts scale and 0.08 for the non addicts scale. Which I thought at first meant that a score of 6 is three times more likely to belong to a non addict than an addict.
But it's better than that, of course. Because we are presumably dealing with very different size data sets. Assuming, generously, that there are 90 non addicts in the data for every 10 addicts (a 10% addiction rate). If there were 1000 participants, then there would be 900 non addicts, and 8% of them would score 6. Say 72 of them. Of the 100 addicts, 2.5% would score 6. Say 3.
So a random person comes along and takes this test, and scores 6. Does the website helpfully explain that while the results cannot possibly be conclusive there is only a 4% chance that someone scoring 6 is a sex addict?
Oddly enough, No.
You've taken the test and it confirmed your fears. You're probably frightened, confused, and overwhelmed. Where do you go? Whom can you trust?...."
Not a clue. But I know who you can't. Hello, Noted psychologist and author Dr Patrick Carnes.
# posted by Louise H @ 11:31 pm 
