Monday, June 24, 2013
The standardization of scores must be done in such a way that we can use one table for all of the normal distributions. This is done by considering how many standard deviations a data value lies from the mean. This makes it to compare a value from one normal distribution to a value from another normal distribution.
For example, suppose Frank and Tom are taking a history course. Frank is in the 8 o'clock class and Tom is in the 10 o'clock class. Since each class has a large number of students, the scores on the final exam each follow a normal distribution. Frank's section had a class average of 75 and his score was a 83. In Tom's section, the average was a 68 and Tom scored a 76. When comparing scores they both felt good that they scored 8 points higher than the average. But which one did better with respect to the rest of the students in the class?
Suppose a graph of the data shows that Frank scored higher than most of the other students in his class, but Tom's score places him more in the middle of the upper half. Then you know that Frank's score is clearly much better with respect to the students in his class than Tom's score is. But how can you tell if the distributions are not graphically represented?
You have to change the raw scores to z scores to see how many standard deviations from the mean each score lies. Suppose the standard deviation of scores in Frank's section is 5 and the standard deviation of scores in Tom's section is 6. The z score is the sample value minus the mean value divided by the standard deviation. Therefore, the z score for Frank is 8/5 = 1.6, and the z score for Tom is 8/6 = 1.33. It's clear that Frank had a better score relative to the class since his score fell 1.6 standard deviations about the mean and Tom's score was only 1.33 standard deviations about the mean.
Note that when standardizing the scores, the mean of the original distribution is always zero, which is logical since the mean lies zero standard deviations from itself. Any value above the mean has a positive z score, and any value lower than the mean has a negative z score.
This guide should help anyone understand how to change a raw score to a z score, which makes comparing scores from different normal distributions easy.