Juryjone,
----
Whew. As I told you, it takes very little in the way of statistics for me to get out of my depth. If you haven't "dumbed it down" as much as you could already, could you explain that in layman's terms?
----
A z-score is something like z = (x-ave of x's)/standard deviation of the x's. It is basically a 'standardized score'. -just a score put on another scale, like Farenheit to Kelvin.
The idea is that while x may have been on a lbs scale, or a height scale, or on a temperature scale, x may have been on another scale in a different analysis, so the scores are standardized so we can compare scores.
Based on the bell curve, a z greater than about 2 (or less than about -2, because the curve is symmetric) is deemed 'significant' at the 5% level.
The z's can take on any value from, say, -10 to 10, where the more positive or more negative the number is, the more 'significant' the result is. Most z's take on values between -3 and 3.
"At a 5% signifigance level' means that if there is no paranormal stuff going on, we'd expect a z this large only 5% of the time or less. So if we get a very very large z, this is evidence against 'no paranormal stuff going on'.
A meta-analysis combines these z's from many experiments. (There are conditions: like they have to be similar experiments where things were measured the same or similar way)
A typical meta-analysis combination of the z's is:
R = squareroot
*average of z's,
where n is the number of z's you are combining.
(which is mathematically the same as (z1+z2+z3+...+zn)/sqrt
, which is what I wrote in the previous message)
We'd be interested in seeing if this R is greater than 2 or less than -2. If it was, the results would be 'significant'.
A z-score of 1 is not significant at the 5% level. But, if all of the z's are mostly of the same sign (+ or -) and are combined, the result is that R could be significant.
Which makes sense because individually they don't mean much, but they are all positive 'effects'.
