In section 7.1, we analyzed normal distributions and specific situations in which analysis was done for data which followed the 68-95-99.7 rule exactly. The truth of the matter is that most situations require us to answer questions that do not reference exact whole numbers of standard deviations above or below the mean. What if we asked a student what their actual score would be if they were in the top 10% of ACT test takers? We need a tool to help us deal with these types of situations.
Our first tool will be the z-score formula. The z-score is a measure of how many standard deviations above or below the mean a particular value is. If a z-score is negative, the result is below the mean and if it is positive, the result is above the mean. For example, if the ACT mathematics exam scores are normally distributed with a mean of 18 and a standard deviation of 6, then an ACT score of 30 would be equivalent to a z-score of 2 because 30 would be 2 standard deviations above the mean.
The formula below gives a quick way to calculate z-scores. In the formula, 'x' is the observation,
is the mean of the distribution,
and is the standard deviation for the distribution.
Example 1
Suppose the mean length of the hair of 10th grade girls is 10 inches with a standard deviation of 4 inches. What would be the z-score for hair length for a 10th grade girl whose hair is 16 inches long and what does it mean in terms of the normal curve?
Solution
It is often a good idea to draw a sketch for these sorts of situations so we can visualize what is happening.
Because 16 is located between 1 and 2 standard deviations above the mean, we expect a z-score between 1 and 2. Use the formula
to calculate the z-score. Our observation, x, is 16 inches while the mean is
the mean is 10 inches and the standard deviation is = 4 inches. or z = 1.5. This tells us that a hair length of 16 inches will be 1.5 standard deviations above the mean.
Example 2
Suppose that the z-score for a particular 10th grade girl's hair length is z = -1.25. What is the length of the girl's hair?
Solution
We will use the z-score formula to find our answer.
The length of the hair for this girl would be 5 inches.
Example 3
Suppose a student can either submit only their SAT score or their ACT score to a particular college. Suppose their SAT score was 620 and that the SAT has a mean of 500 and a standard deviation of 100. Suppose also that the same student scored a 25 of their ACT exam and that the ACT exam has a mean of 18 and a standard deviation of 6. Which score should the student submit?
Solution
Looking at the diagram below, it is not exactly clear which score is better. They appear to be quite similar and we will need to do some calculations to make a distinction.
Calculate the z-score for each exam, for the SAT, for the ACT. Since the z-score is higher on the SAT, the student should submit the SAT exam score.
In order to understand how to apply z-scores beyond what we have already done, we must first understand percentiles. A percentile is a marker on a normal curve such that the marker is greater than or equal to that percentage of results. For example, suppose you are at the 30th percentile for how fast you type. This means that you can type faster than 30% of all people. The percentile can also be thought of as the percent of area to the left of its marker. The graphic below shows where the 30th percentile is located. The shaded area to the left of the marker represents 30% of the normal curve.
It is very common for colleges and universities to use percentiles for entrance criteria. For example, a rather elite university might require that you score at the 90th percentile or higher on your ACT exam to be considered for admissions. Doctors often use percentiles to track the growth of babies. For example, can you picture what a baby would look like that is at the 70th percentile for weight and the 25th percentile for length?
Now we must ask what percentiles have to do with z-scores. Find the Normal Distribution Table in Appendix A, Part 2 of the website. Let's examine the z-score of -1.25 from Example 2. Find the z-value of -1.2 and then go over until you are under the 0.05 column. A partial table is given in Figure 7.3 below and the value in the cell we are looking for is bold and underlined. The value of 0.1056 can be interpreted as a percentile. This means that the girl in Example 2 has hair that is longer than 10.56% of all girls. In other words, she is at about the 10th or 11th percentile for hair length for 10th grade girls.
Example 4
At what percentile for hair length is a 10th grade girl if her hair is 17 inches long?
Solution
Start by determining her z-score which would be
We now go to the Normal Distribution Table in Appendix A, Part 2 of the book. We go across the row with z=1.7 until we are under 0.05. This gives a value of 0.9599. This tells us the girl is at about the 96th percentile for hair length. In other words, this girl's hair is longer than 96% of all 10th grade girls.
While it is nice to find percentiles for certain situations, we are often asked for the percentage of results that are between two given parameters or above a given parameter. For example, we might be asked to find the percentage of all 10th grade girls that have hair lengths between 8 inches and 15 inches long. To find these types of results, we often must do multiple z-score calculations and some addition or subtraction.
Example 5
Suppose the weights of adult males of a particular species of whale are distributed normally with a mean of 11,600 pounds and a standard deviation of 640 pounds.
a) What percent of these adult male whales will weigh between 11,000 and 12,000 pounds?
b) What percent of these adult male whales will weigh more than 12,000 pounds?
Solution
a) Begin by finding the z-scores for both of the weights given.
For z=-0.9375, our Normal Distribution Table in Appendix A, Part 2 gives us a value between 0.1736 and 0.1762. Since -0.9375 is closer to -0.94 than -0.93, we will use a value of 0.174. We get a value between 0.7324 and 0.7357 for z=0.625. We will split the difference on this and use 0.734. All that is left to do now is subtract 0.734 and 0.174 to get 0.56 or about 56% of all adult male whales of this species are between 11,000 and 12,000 pounds. The shaded region in the Figure 7.4 below represents about 56% of the normal curve.
b) Use z=0.625 from part a) to get a value from the table of 0.734. This means that 73.4% of all whales weigh 12,000 pounds or less. Therefore, 100%-73.4%=26.6% of all whales weigh more than 12,000 pounds.
It is also important to note that graphing calculators can be used to quickly solve the types of problems discussed in this section by using the NormalCdf command. Typically, this command requires that four values be entered, the lower bound, the upper bound, the mean, and the standard deviation. In Example 5, we can solve the problem in part a) simply by typing in the command string NormalCdf(11000,12000,11600,640) and obtain the immediate result of 0.5598 or 56%.
Be sure you know how to access this command if you have a graphing calculator. Appendix C has some notes for users of graphing calculators. An online calculator that is very similar to a graphing calculator and gives us the same information can be found at http://wolframalpha.com .
You might also be wondering how to solve a problem using the NormalCdf command if only one parameter is given. Let's revisit Example 4 to see how this works.
Example 6
At what percentile for hair length is a 10th grade girl if her hair is 17 inches long?
Solution
There is only one boundary given in this problem. It is your job to come up with a second boundary. In this case, the percentile we want to calculate is found by finding the percentage of all girls whose hair is 17 inches or less. We will use a lower bound of -100 and an upper bound of 17. We use -100 simply because we are confident that we will not find any results any further left than this. Typically, choose your missing parameter as being so extreme that it will not be even in the realm of possible results. NormalCdf(-100,17,10,4)=0.9599 so the length of the girl's hair is at about the 96th percentile.