Finding out what’s special – Z-Scores
Z or standard scores are a good way to figure out what is special. Let’s say you have election results and want to find places that are interesting to report on. One thing you can do is figure out where a party has done exceptionally well. Z-Scores are great for this.
The Z-Score of a measurement is calculated as
(x – mean)/standard deviation.
The Z-Score gives you a value’s distance in standard deviations from the mean. You can now set arbitrary limits and say: places where the votes with a Z-score higher than 2 are interesting because this means that extraordinarily many people voted for a specific party. A Z-score below -2 is also interesting because exceptionally few people voted. You can make this a little finer by looking at Z-scores for counties or regions (often there are regional differences and so on). Remember – 95.45 percent of measures fall into 2 standard deviations from the mean (if you have normally distrubuted data) this means: there is less than a 5 percent change for the Z score to be higher (or lower) than 2. 5% makes you pretty stand out. Not enough: 3 Standard deviations give you less than 1% of chance (99.73% of points are within 3 standard deviations from the mean in a normal distribution).
Overall, the Z-Score is a handy addition to your toolbox for figuring out what data values are different.
