-assuming they belong there....
Just a bell shaped curve that turns up a lot when dealing with large numbers in a sample
--> "What kind of curve did you get"
"Oh, just a normal one"
- e.g., height: few really small, few really tall, most lumped in middle
- curve never touches the base line (can't have 0 height)
Bell shaped curve happens to have a lot of handy qualities statistically
- 34% of population between 1SD above & below the mean
- 14% between +1 and +2 or between -1 and -2;
- only about 2% above +2 or below -2
- almost everybody above -3 and below +3 (1 in 10,000 case outside)
- so if sample is below 10,000, you don't have anyone out there...
- and of course, as with any symmetrical curve, mean and median
are the same (and the mode, unless its bimodal)
- so in normal curve, half are above average, half below
Percentile
- point on distribution at or below which you find given percentage of
individuals
- 50th percentile means 50% got that score or less --i.e., divides in half
-
Quartiles
- divides frequency distributions into equal fourths
- first (called Q ) is 25th percentile, 2nd = 50th percentile
1
third = 75th percentile
- can't say "in Q1", has to be "at or above Q1"
Deciles
same deal, only into tenths: 10% below first decile, 20 below second, etc
Stanines
- normal curve divided into 9 equal parts
- each has an interval of 1/2 standard deviation
- easy to calculate, gives you single digit score
- forces any distribution (no matter how badly skewed) into a normal curve
- i.e., if everybody got between 85 and 90% in a course, could still
rank everyone from top to bottom by forcing onto a normal curve
and assigning a stanine score
- requires that you know that the population is normally distributed
- AND requires that the sample you are working with is not skewed
- it is unlikely that you will be in a situation where it will be
appropriate for you to use stanines