Lesson 5: Standard Scores and Normal Distribution

5.1 Comparing Scores across Distributions

The picture above does a great job of introducing the concept of standard scores and why
comparing across distributions is tricky. The gentleman was under the impression the
numbers the lady gave him meant something other than what she meant, that is, how the
gentleman interpreted 42 was not what the lady meant as 42. Thus, the cartoon illustrates
that when comparing across distributions, you cannot rely on raw scores; rather, raw data
must be converted into a standard measure that will be understandable in any distribution.

To understand why comparing across distributions is problematic, consider the following

situation: You have a friend, Stu, who thinks he is smarter than you. Both you and Stu take
the same 19 Century American Literature course from the same professor, in the same
th

semester, but in different class sections. You have a test in that course on the same day and
the content of the test is exactly the same. On the test, Stu obtains 85 out of 100 points and
you obtain 80 out of 100 points. Of course, Stu starts bragging and boasting about his
19th Century American Literature intelligence. But who really did better on this test?

It’s true your score of 80 was less than Stu’s score of 85, but what happens when you consider
information about each class’ test score distribution? Say each test was on the same material
and included the same content, but the makeup of each test was slightly different (e.g., more
multiple choice questions for one section, slightly differently-worded questions). Indeed,
maybe the classes were a little different since there are different students in each section and
the professor may adjust change lessons accordingly. That said, assume your class’ mean
test grade was M = 75 points and Stu’s class mean was M = 80 points. Now who did better?
When you take the mean of each class into account, it seems you and Stu performed equally,
because both scores were 5 points above your respective class mean. However, because the
means of each class are different, this suggests the underlying test score distributions in each
class were also different; hence, a 5 point difference in one class may be different than a 5
point difference in the other. Thus, the 5 point differences must be considered relative to the
class distribution.

So, who did better, you or Stu? What you need is a ‘standard measure’ to compare your
scores across the two class distributions. Fortunately, we already know of such a measure:
the standard deviation. Remember, the standard deviation is the average difference between
a score (x) and the mean (M) of a distribution. Although the value of the standard deviation
changes depending on the data in the distribution, a standard deviation means the same thing
in any distribution. For example, one standard deviation from the mean means the same thing
regardless of the underlying distribution. Similarly, knowing a score is two standard deviations
away from the mean would mean the same thing in any distribution. Thus, the standard
deviation measures the distance between a score and the mean of a distribution and that
distance can be used to compare across different distributions, because standard deviations
have a universal meaning.

When you measure the number of standard deviations between your test score and the mean
of your class, you can compare that difference, which is in numbers of standard deviations,
to the number of standard deviations between Stu’s test score and the mean of his class. The
question is how do you do this? The measure of the number of standard deviations between
any raw score and the mean of a distribution is a standard score, or z-Score.

5.2 Standard (z) Scores

A standard score (z-Score) measures the number of standard deviations between a score
(X) and the mean (M or μ). To calculate a standard (z) score for a score, you need the mean
of the distribution (M or μ), the standard deviation (s or σ), and a raw score (X). The formula
for calculating a standard score is:

And in a sample is:

Both formulas are functionally equivalent: subtract the mean from a score (X) and divide that
difference by the standard deviation. The ’z’ comes from a standard scores when it is
calculated from a raw score in a normal distribution (more on this later).

Back to you and Stu: Assume the standard deviation for your class is s = 1 and the standard
deviation for Stu’s class is s = 5. Remember, your score was X = 80 and Stu’s was X = 85.
Your class' mean was M = 75 and Stu’s class mean was M = 80. We have:

The difference between each raw score and the mean of each class is identical (difference =
5), but Stu’s standard score is less than yours (z = 1 vs. z = 5). The reason has to do with
the difference in the variability among the test scores between classes. The standard deviation
was smaller in your class than Stu’s class, and a smaller standard deviation suggests less
variability among scores in your class compared to Stu’s. Less variability indicates the scores
in your class were, on average, closer to the mean, compared to Stu’s class where the higher
variability indicates greater differences from the mean. Stated differently, the higher standard
deviation in Stu’s class suggests a greater dispersion of scores around the mean. Because
most scores in your class were similar to the mean, deviations from your class’ mean are
unlikely, but because there is higher variability in Stu’s class deviations from the mean are
not surprising.

Standard scores can be used to compare test performance across the two class distributions.
Because your score is farther away from the mean than Stu’s score, in standard deviations
you did better on the test. Thus standard scores provide a way to compare across
distributions. In short, because raw scores mean different things in different distributions, it
is appropriate to convert scores (X) into standard (z) scores when comparing across
distributions.

5.3 Characteristics of Standard (z) Scores

There are several noteworthy characteristics of standard scores. First, standard scores can be
used to compare across distributions as was done in the example in Section 6.2. Again, when
comparing across distributions, unless the mean and standard deviation are identical in each
distribution or if the raw scores come from the same distribution, it is inappropriate to
compare scores; you should use standard scores.

Second, larger standard scores indicate larger deviations (distances) from the mean. Thus,
the larger the standard score, the farther the raw score (X) is from the mean. And, of course,
a standard score of zero (z = 0) indicates a score lies at the mean.

Third, standard scores are defined as the distance between a score and its mean in standard
deviations; thus, your z-score from Section 6.2 of z = 5 indicates your score of X = 80 is five
standard deviations above your class’ mean, and Stu’s z-score of z = 1 indicates his score of
X = 85 is one standard deviation above his class’ mean. Similarly, a z-Score of z = -1.5
indicates a score is one and one-half standard deviations below the mean.

Fourth, standard scores can be positive or negative. Positive z-scores indicate the score was
greater than the mean and negative z-scores indicate the score was less than the mean.
Importantly, the sign (+/-) does not say anything about magnitude of z-score. For example,
a z-score of -2.0 is larger than a z-score of +1.0, because, z = -2.0 is twice the distance from
the mean compared to z = +1.0. Thus, the sign (+/-) of a z-score indicates the direction a
score from the mean; the absolute value indicates magnitude.

Finally, if you calculate the mean, variance and standard deviation of a distribution of z-scores
calculated from a set of scores, the mean of the z-distribution will be zero and the variance
and standard deviation will be equal to one:

X (X-X) (X-X)2 z (z-z) (z-z)2

4 1 1 1.118 1.118 1.25
4 1 1 1.118 1.118 1.25
3 0 0 0 0.000 0
2 -1 1 -1.118 -1.118 1.25
2 -1 1 -1.118 -1.118 1.25
∑X=15 SS = 4 ∑z = 0 SS = 5
Z

M=3 s = 0.8
2 sz2=1
 s = 0.894 sz=1

5.4 Using the Mean, Standard Deviation, and Z to find Raw

Scores
You can use the standard deviation to determine the distance between a score and the mean,
but what if you want to determine the score that would be a certain number of standard
deviations above or below a mean? That is, say we have a distribution of 19 Century th

American Literature test scores with mean M = 70 and standard deviation s = 10. Assume
that any score 2.5 standard deviations or more above the mean earns an A; that is, the
minimum score for an A is associated with z = 2.5. So, we want to know the score (X) that is
exactly 2.5 standard deviations above the mean. How can we do this?

Remember a z-score measures the number of standard deviations a raw score is away from
a mean. In the example above we are looking for the raw score associated with z = 2.5. To
calculate this raw score we use one of the two following equations, both of which are used to
determine the raw score associated with a z-score, given we know the mean and standard
deviation: X = zσ + μ And in a sample use: X = zs + M.
Both formulas are functionally equivalent, because both formulas determine a raw score (X)
by adding the mean of a distribution to the product of a z-score and the standard deviation
of a distribution. From the example above, the minimum raw score needed for an A can be
determined by insert the class’ mean (M = 70), standard deviation (s = 10), and z-score (z
= 2.5) into the formula above: X = (2.5)(10) + 70 = 95.

A score of X = 95 is exactly 2.5 standard deviations above the mean, so the minimum grade
needed for an A is 95. What about a score that lies 1.25 standard deviations below the mean
(z = -1.25)? Maybe this is the minimum grade needed to pass the test. Determining this score
is the same as before, you just need to be sure to take the negative z-score into account: X
= (-1.25)(10) + 70 = 57.5. Thus, a score of 57.5 is exactly 1.25 standard deviations below
the mean of 70, and is the minimum score needed to pass the test.

5.5 Standard Scores and the Normal Distribution

When you have a population or a sufficiently large sample of normally distributed data, a
certain proportion of scores will fall between the mean and any given number of standard
deviations from the mean. Specifically, the proportion of scores between the mean (μ or M)
and one standard deviation from the mean is approximately .3413 (i.e., 34.13% of scores fall
between the mean and one standard deviation from the mean, above or below). Similarly,
the proportion of scores between one standard deviation and two standard deviations from
the mean is approximately .1359. Incidentally, the proportion of scores between one standard
deviation below the mean and one standard deviation above the mean is approximately
.6826; hence, most scores in a normally distributed set of data fall right around the mean
(i.e., central tendency).

These proportions can be seen in the general graph of the normal distribution, below. I should
note this is a theoretical normal distribution centered on a mean and is symmetrical with half
the scores greater than the mean and the half less than the mean. Remember, in a normal
distribution most scores lie at the hump in the middle and fewer scores lie in the tails.
The height of the curve above any point in the x-axis reflects the probability of observing that
particular score (X); hence, at any point along the x-axis, a certain proportion of scores are
above and below that point. The higher the height of the curve, the more probable that score
is likely to be observed and the greater the proportion of the distribution is made up of that
score. As you can see in the graph above, as the distance between a score (X) and the mean
increases, the height of the curve and the probability of observing that score decreases.

Along the x-axis, marks are given for the number of Standard Deviations from The Mean,
from four standard deviations below the mean (-4σ) to four standard deviations above the
mean (+4σ). Marks are also given to indicate the Z Scores relative to the mean from z = -
4.0 to z = +4.0. (Don’t worry about the Cumulative % and T Scores) As you can see, as you
move farther and farther away from the mean in standard deviations, or z-Scores, the
probability of being that far away become less and less. This fact about the normal distribution
introduces two key points:

First, because the proportions/probabilities in the figure above are associated with specific
numbers of standard deviations from the mean, and because standard deviations are a
measure that can be used in any set of data, the proportions/probabilities associated with z-
scores are universal and occur in any normally distributed data. For example, IQ scores are
normally distributed with μ = 100 and σ = 16. This means the proportion of scores between
an IQ score of X = 84 (μ - 1σ) and X = 116 (μ + 1σ) is approximately 0.6826. If we assume
scores on the Beck Depression Inventory (BDI) are normally distributed with μ = 15 and σ =
5, the proportion of scores between a BDI score of X = 10 (μ - 1σ) and X = 20 (μ + 1σ) is
also approximately 0.6826. Thus, regardless of the underlying data, as long as I is normally
distributed these proportions/probabilities will be found for any z-score.

Second, by taking into account the mean and standard deviation of a distribution, we can
calculate a z-score for any raw score and we can use the information about the normal
distribution above to determine proportions and probabilities associated with that z-score,
which we do in the next section.

5.6 Probabilities and Proportions in the Normal

Distribution
Table 1 in Appendix A presents the is the Standard Normal (z) Table, which can be used
to look up proportions of scores above or below z-Score within a normal distribution. A portion
of Table 1 is presented below. There are three columns in Table 1, the first column lists z-
Scores that increase from 0.00 to 4.00, mainly in increments of 0.01, and the second and
third columns list proportions of scores under the normal curve that are between the z-score
and the mean (column 2) and beyond that z-score and into the tail of the distribution (column
3).

Column Column
Column 2 Column 3 Column 2 Column 3
1 1
p (0 < x ≤ +z) p (x ≥ +z) p (0 < x ≤ +z) p (x ≥ +z)
z or or z or or
p (0 > x ≥ -z) p (x ≤ -z) p (0 > x ≥ -z) p (x ≤ -z)
1.00 0.3413 0.1587 1.60 0.4452 0.0548
1.01 0.3438 0.1562 1.61 0.4463 0.0537
1.02 0.3461 0.1539 1.62 0.4474 0.0526
1.03 0.3485 0.1515 1.63 0.4484 0.0516
1.04 0.3508 0.1492 1.64 0.4495 0.0505
1.05 0.3531 0.1469 1.65 0.4505 0.0495
1.06 0.3554 0.1446 1.66 0.4515 0.0485
1.07 0.3577 0.1423 1.67 0.4525 0.0475
1.08 0.3599 0.1401 1.68 0.4535 0.0465
1.09 0.3621 0.1379 1.69 0.4545 0.0455
1.10 0.3643 0.1357 1.70 0.4554 0.0446

Notice there are no negative z-Scores. This is because the sign (+/-) of a z-Score does not
tell you anything the magnitude of a z-score r its distance from the mean; the sign only tells
you the direction of the z-score relative to the mean. Thus, z = 1.25 is equal in distance to z
= -1.25 and the proportions under the normal curve associated with a positive z-score will be
the same as the proportions under the normal curve associated with a negative z-score of the
same absolute value. If you have a negative z-score, just look up a positive z-score of the
same absolute value and use those proportions in Table 1.
Before going into detail about what information columns 2 and 3 provide, let’s set up an
example. We have a normally distributed set of 19 Century American Literature test scores
th

with μ = 80 and σ = 10. We calculate the z-score for the raw score of X = 92 and find it is z
= 1.2. We now want to know the proportion of scores that lie between z = 1.2 and the mean
of the distribution. Another way to conceptualize this question is we want to know the
probability of obtaining a score between the mean and z = 1.2. Thus, we want to know p(μ
≤ X ≤ 92). The section of Table 1 that includes z = 1.2 is reproduced below with the area
associated with z = 1.2 highlighted in yellow:

Column Column
Column 2 Column 3 Column 2 Column 3
1 1
p (0 < x ≤ +z) p (x ≥ +z) p (0 < x ≤ +z) p (x ≥ +z)
z or or z or or
p (0 > x ≥ -z) p (x ≤ -z) p (0 > x ≥ -z) p (x ≤ -z)
1.15 0.3749 0.1251 1.75 0.4599 0.0401
1.16 0.3770 0.1230 1.76 0.4608 0.0392
1.17 0.3790 0.1210 1.77 0.4616 0.0384
1.18 0.3810 0.1190 1.78 0.4625 0.0375
1.19 0.3830 0.1170 1.79 0.4633 0.0367
1.20 0.3849 0.1151 1.80 0.4641 0.0359
1.21 0.3869 0.1131 1.81 0.4649 0.0351
1.22 0.3888 0.1112 1.82 0.4656 0.0344
1.23 0.3907 0.1093 1.83 0.4664 0.0336
1.24 0.3925 0.1075 1.84 0.4671 0.0329

Column 2 in the Table provides the proportion of scores between the mean a given z-score;
thus, the proportion of scores between the mean of μ = 80 and the raw score of X = 92 (z =
1.2) is 0.3849. This would be the same proportion of scores between the mean and z = -1.2,
because the proportion of scores between the mean and a positive z-score is equal to the
proportion of scores between the mean and a negative z-score of the same magnitude.

We can also use Table 1 to determine the proportion of scores beyond a given z-score, that
is, the proportion of scores that fall in the tail of normally distributed. Another way to
conceptualize this question is, what is the probability of having score greater than z = 1.2;
thus, what is p(1.2 ≤ X). This proportion comes from Column 3 in the table above. For z =
1.2, the proportion of scores that fall in the tail 0.1151; hence, the probability of having a
score greater than X = 92 (z = 1.2) is p = 0.1151. Note, the proportion of scores beyond z =
1.2 is equal to the proportion of scores in the tail for z = -1.2. Importantly, if the z-score is
positive, this is the proportion of scores greater than the score (X), but if the z-score is
negative, this proportion of scores less than the score.

The proportion of scores between the mean and any z-score (Column 2) to the proportion of
scores that lie beyond that z-score (Column 2) add to 0.5000. This is because the proportion
of scores above the mean in a normal distribution is always 0.5 and the proportion below the
mean is always 0.5. This is so, because in a normal distribution the mean is equal to the
median and there are always 50% of the scores above the median and 50% of the scores
below the median.

Knowing this, you can use Table 1 to determine the probability of obtaining a score greater
than or less than a given score (X). For example, with the same normally distributed set of
19 Century American Literature test scores (μ = 80; σ = 10), say we want to determine the
th

probability of obtaining a score less than or equal to a raw score of X = 75. The z-Score for
75 in this distribution is z = -0.50. Look up z = 0.50 in Table 1, because there are no negative
z-Values in the table. We are interested in the probability of obtaining a score less than or
equal to X = 75 (z = -0.5), so we use the proportion in Column 3, which tells us that the
probability of obtaining a score of 75 or less is p (X ≤ 75) = 0.3085. We use the proportion
from Column 3 in Table 1, because that is the probability of being at that z-score or lower.

What about finding the probability of obtaining a score of X = 75 or more? In this case you
use the same z-Score of z = -0.50 (again, use z = 0.50 in Table 1). There is no single
proportion in Table 1 that provides you the answer, so what you do is find the probability of
a score between the mean and z = 0.50, which is 0.1915 from Column 2, and add 0.5000.
Adding 0.5000 takes into account the upper half of the distribution; that is, the probability of
obtaining X = 75 or more is equal to the probability of obtaining a score between that raw
score (X = 75) and the mean (μ = 80) added to the probability of obtaining a score in the
other half of the distribution (above the mean). So, the probability of obtaining a score of 75
or greater is 0.6915.

How about the probability of obtaining a score of 90 or less? Again, calculate the z-score,
which is z = 1.00. In this case, we need the probability below z = 1.00, which is equal to the
area between the mean of the normal distribution and z = 1.00 (0.3413 from Column 2)
added to the total area below the mean (0.5000). So, the probability of a score below 90 is
0.3413 + 0.5000 = 0.8413.

What if you wanted to calculate the probability of obtaining a score between two values?
There are two ways to do this: (1) When both values fall on the same side of the mean, and
(2) when one values falls above the mean and the other value falls below the mean.

From the same distribution of 19 Century American Literature test scores with μ = 80 and σ
th

= 10, we want to know the probability of obtaining a score between 65 and 85. First, calculate
the z-Scores for each raw score, which are z = -1.50 for 65 and z = 0.50 for 85. Next, look
up the probability between each z-score and the mean in Column 2, which is 0.4332 for z =
-1.50 and 0.1915 for z = 0.50. Because each score falls on a different side of the mean (one
z-score was positive the other was negative), add these probabilities. So, the probability of
obtaining a score between 65 and 85 is = 0.4332 + 0.1915 = 0.6247.

Say we want to determine the probability of obtaining a score between 85 and 90, which are
on the same side of the mean. First, calculate the z-scores for each raw score, which are z =
0.50 for 85 and z = 1.00 for 90. Next, look up the probability between each z-Score and the
mean in Column 2, which is 0.1915 for z = 0.50 and 0.3413 for z = 1.00. Because both scores
are on the same side of the mean, subtract the smaller probability from the larger probability,
which will leave you with the probability between 85 and 90. So, the probability of obtaining
a score between 85 and 90 is 0.3413 - 0.1915 = 0.1498 (remember, probabilities cannot be
negative).

The same procedure is used if both scores are less than the mean. For example, we want to
know the probability of obtaining a score between 65 and 75, which are less than μ = 80.
First, calculate the z-scores for each score (z = -1.50 for 65 and z = -0.50 for 75). Next, look
up the probability between each z-score and the mean in Column 2 (0.4332 for z = -1.50 and
0.1915 for z = -0.50). Because both values are less than the mean, subtract the smaller
probability from the larger probability. So, the probability of a score between 65 and 75 is
0.4332 - 0.1915 = 0.2417. Important: When calculating the probability between two scores
that are on the same side of the mean, you will always subtract the smaller probability from
the larger probability. But, when calculating the probability between two scores that are on
different sides of the mean, add the two probabilities together.