|A short statistical evaluation with a crack of the American SPSS program of the zero and one project.|
|Introduction: The zero and one project is very simple to
understand, we have so called 'zero' days and we have so called 'one'
On the zero days it is advised or expected that we have less US slime
killed in Iraq.
Under observation were 94 days ranging from 24 March to (and with) 25 July 2007; 94 days in total.
Battlefield considerations indicate it could be handy if the Iraqis
know beforehand if certain days are a 'zero' or a 'one' kind of day.
To give you an idea about how stuff works we have a 'cut and paste' from the main page:
My thanks go once more to the Iraqis who have supported this project,
without there help this would have been impossible.
Lets take a look in detail now this introduction runs at it's end.
|The look in detail:
Since most of my readers are not familiar with statistical testing anyway and the SPSS program in particular it is most wise to make a few screenshots so the info needed to get out will indeed get out.
In order to understand what the next screenshots mean you only have to
understand the meaning of two variables, the variables are named SLIME and
MBDper. Both are simple to understand, when for example SLIME has value 5
on a certain day this only means we had five filled US coffins on a
particular day, so the SLIME vafriable traces the daily US (and coalition)
death toll. Simple or not?
SLIME = daily US death toll &
MBDper = the period we are in.
We start this series of screenshots with a very simple one, all days are combined and we only put on the X-axis the daily US death toll and on the Y-axis we put the number of days into our time basis of just 94 days.
For example you see that during four days we had 8 soldier slime
Now how does such a split up look on the so called ´zero´ days you
Where on the ´one´ days statistics break down to
Needless to say there is a great difference between the last two graphs so it is wise to study further what differences there are in particular.
Therefore we take a look at the so called ´descriptives´ inside the SPSS package and we arrive at the next screen shot that says
Ah, now we see daily US slime killed is 3.09 in the zero days while it runs at 4.22 in the one days (remark we had 94 days under investigation)
So is there any kind of significance found between these both averages or is the just pure coincidence...
For this to estimate we only have to apply standard scientific knowledge and this knowledge says
H0 There is no difference whatsoever, versus
Ha We might have a problem because the differences are just too big.
Let me spare you the technical details but the next screen shot simply says
And for the statistical trained eye you only look at the two tailed significance in the upper row, divide it by two to get 0.057 / 2 equals 0,0285.
Therefore my fellow US statisticians will jump to the conclusin that I am in the right where Dubya is telling stupidity.
Lets proceed with so called correlations, when we calculate correlations we see once more this 0,057 number propping up.
Here is the screen shot
And why not take a look at some box plots (as far as I know it these are 95% threshold boxplots)
And lets end this update with something I am not familiar with, I only publish this because I am unsure about the real facts. Let me share my worries with you± I thought I had the normal distribution in the back of my car but now a more rigid analysis with help of the Komogorov testing as found in the SPSS package using the explore options has made me more worried...
Lets just place into fact that normality of these battlefield date is still a matter of discussion...
That was it, may be later to be updated.