SUBSIM Radio Room Forums - View Single Post

Skybird · 08-23-06, 05:15 PM

It's ten years ago, and so I do not have it all on my mind anymore. But we even had very different statistics for the same time period about the same country!

BTW, your graphs only describe a correlation, somewhat (not really, but you emphasize the link between two variables without further elaboratin it). Every academic who is trained in statistics will tell you that a correlation never - NEVER - tells you something about a causal link (nor does the display of just two graphs). A correlation coefficient (or the two graphs shown) only tells you something about to what degree the two variables tend to show "linked" values, for whatever a reason (there could be third and more variables involved). So, WHY they do that is a completely different story. In your graphic it means that the fact that the two graphs in your interpretation mirror each other's meaning, does not autpmatically mean that the one variable (number of death sentences) is causing the result of the other (crime rate). Like if you find a correlation between hair colour and size of shoes does not mean that the colour of your hair has an influence on the size of your feet. the drop in crime rate could be caused by very different things, and the graph of executions simply is a coincidence. You need far more statistical analysis and an elaboration on the raw data to come to a more meaningful conclusion.

The public is often fooled by simplified statistics, to get it into the direction an interested party wants it to move at.

In other words: that simple graphic - for the time being means nothing. It could be that some defender of death penalty just arranged it while ignorring the statistical background analysis, knowing that it would catch people's eyes and that most would willingly interpret it the way you just did yourself. Even if the counting results are correct - it still does not mean anything. It is bad statistical procedure, and bad academical procedure. It could be very different. Maybe more police personnell (just an example). Less poverty leading to less robberies with murder. Less alcohol or less love affairs leading to murderings commited as crimes of passions. Or a love&justice epidemic brought out. Who knows...

As my old statistics prof time and again was preaching us: "A statistical mean value is absolutely worthless if given without a couple of additional discriptive values, such as variance, and the like." Right he was.

08-23-06, 05:15 PM	#68
Skybird Soaring Join Date: Sep 2001 Location: the mental asylum named Germany Posts: 42,629 Downloads: 10 Uploads: 0	It's ten years ago, and so I do not have it all on my mind anymore. But we even had very different statistics for the same time period about the same country! BTW, your graphs only describe a correlation, somewhat (not really, but you emphasize the link between two variables without further elaboratin it). Every academic who is trained in statistics will tell you that a correlation never - NEVER - tells you something about a causal link (nor does the display of just two graphs). A correlation coefficient (or the two graphs shown) only tells you something about to what degree the two variables tend to show "linked" values, for whatever a reason (there could be third and more variables involved). So, WHY they do that is a completely different story. In your graphic it means that the fact that the two graphs in your interpretation mirror each other's meaning, does not autpmatically mean that the one variable (number of death sentences) is causing the result of the other (crime rate). Like if you find a correlation between hair colour and size of shoes does not mean that the colour of your hair has an influence on the size of your feet. the drop in crime rate could be caused by very different things, and the graph of executions simply is a coincidence. You need far more statistical analysis and an elaboration on the raw data to come to a more meaningful conclusion. The public is often fooled by simplified statistics, to get it into the direction an interested party wants it to move at. In other words: that simple graphic - for the time being means nothing. It could be that some defender of death penalty just arranged it while ignorring the statistical background analysis, knowing that it would catch people's eyes and that most would willingly interpret it the way you just did yourself. Even if the counting results are correct - it still does not mean anything. It is bad statistical procedure, and bad academical procedure. It could be very different. Maybe more police personnell (just an example). Less poverty leading to less robberies with murder. Less alcohol or less love affairs leading to murderings commited as crimes of passions. Or a love&justice epidemic brought out. Who knows... As my old statistics prof time and again was preaching us: "A statistical mean value is absolutely worthless if given without a couple of additional discriptive values, such as variance, and the like." Right he was. __________________ If you feel nuts, consult an expert. Last edited by Skybird; 08-23-06 at 05:25 PM.