MEASURES OF VARIABILITY
Range
A very simple measure of variability is to get the score difference between the highest
and the lowest score and this measure is called the range of the distribution. If in a reading test
for example, the highest score is 95 and the lowest is 45, the range is 40. However, the range
depends only upon the 2 extreme scores in the total group. This makes this measure a very
unreliable because it can be changed a good bit by the inclusion or omission of a single extreme
case. The example below illustrates that the range of a set of scores is affected by a single
extreme score. The Range for group 1 is 40 while the range for group 2 is 70.
Group 1 45 50 76 77 80 81 90 95
Group 2 25 50 76 77 80 81 90 95
Think about this!
Can we use the range to compare two groups relative performance? Does a higher value of
range indicate lower performance? Why? Why not?
Semi-Interquartile Range or Quartile
Another measure of variability is the range of scores that includes a specified part of the
total group – usually the middle fifty percent. The middle fifty percent of the group are scores
lying between the 25th and 75th percentiles. The 25th(Q1)and 75th (Q3) percentiles are called
quartiles since they cut off the bottom quarter and the top quarter of the group respectively. The
score distance between them is called the interquartile range. The statistic that is often reported
as a measure of variability is the semi-quartile range (Q), which is half of the interquartile range.
Q3−Q1
So, Q=
2
Finding Q 1and Q3
Here is a set of score from a class of 50 student. The scores have already been summarized into
a frequency distribution.
X f Cf
95-99 3 50
90-94 4 47
85-89 5 43
80-84 8 38
75-79 6 30 Q3
70-74 10 24
65-69 4 14
60-64 4 10 Q1
55-59 2 6
50-54 0 4
45-49 1 4
40-44 3 3
N
Q1=¿ Q 1 +
i( 4
−cf )
f Q1
Where : Q1= the 25th percentile
LQ1 = lower limit of the 25th percentile class
N = total number of frequencies in the distribution
Cf = cumulative frequency of the median class
fQ1 = frequency of the first quartile (25th percentile)
i= size of the interval of the first quartile
Note: Just like in finding the median of a set of grouped scores, the first step is to get the cumulative
frequency. In the formula, N is divided by 4 (N/4), because we are looking for the score which is at
middle of the lower half of the distribution. Just like the median, the same procedure is applied in
looking for the value of cf, f and the lowest limit.
N
So,
Q1=¿ Q 1 +
i( 4
−cf )
f Q1
= 64.5 + 5 [(50/4 - 10)/4]
= 64.5 + 5 [(12.5-10)/4]
= 64.5 + 5 (2.5/4)
= 64.5 + 5(.63)
= 64.5 + 3.15
= 67.65
Similarly, we use the same formula in finding Q3. The only difference is to get ¾ of N,beacsue we are
looking for the score that lies at the middle of the top half of the distribution.
3N
Q3=¿ Q 3 +
i
4 ( −cf )
f Q3
Where : Q3= the 75th percentile
LQ3 = lower limit of the 75th percentile class
N = total number of frequencies in the distribution
cf= cumulative frequency of the 3rd quartile class
fQ3 = frequency of the third quartile
i= size of the interval of the 3rd quartile class
3N
So,
Q3=¿ Q 3 +
i( 4
−cf )
f Q3
= 79.5 + 5 [ (37.5- 30)/8 ]
= 79.5 + 5 (7.5/8)
= 79.5 + 5 (.94)
= 79.5 + 4.7
= 84.2
84.2−67.65
Therefore: Semi-quartile range Q= =8.275
2
Deciles
The score that divides the distribution into ten equal parts is called the deciles. Just like
Q1 and Q3, we can compute the Decile by determining the number of cases required . For
example, we are looking for the 1 stDecile (D1), then we divide N by 10, the 2N/10 for D 2, 3N/10
for D3 , and so on.
Example: Look for D2
X f Cf
75-77 3 50
72-74 5 47
69-71 2 42
66-68 4 40
63-65 3 36
60-62 0 33
57-59 2 33
54-56 4 31
51-53 6 27
48-50 3 21
45-47 4 18
42-44 6 14 D2
39-41 5 8
36-38 1 3
33-35 2 2
N=50
2ndDecile or D2 = LD2+ i [ (2N/10 – cf)/ f D2 ]
= 41.5 + 3 [(.20x50)-8]/6
= 41.5 + 3 [(10-8)/6 ]
= 41.5 + 1.00
= 42.50
Percentiles
The same procedure may be used when we find the score below which any percentage of
the group falls. These values are called percentiles. The median is the 50th percentile, i.e., the
score below which 50 percent of individuals fall. If we want to find the 40 th percentile, we must
find the score below which 40 percent of the cases fall. Any other percentiles can be found in
the same way. Percentiles have many uses, especially in connection with test norms and
interpretation of scores
Below is an illustration how a particular percentile is computed.
X f cf
75-77 3 50
72-74 5 47
69-71 2 42
66-68 4 40
63-65 3 36
60-62 0 33
57-59 2 33
54-56 4 31
51-53 6 27
48-50 3 21
45-47 4 18
42-44 6 14 P20
39-41 5 8
36-38 1 3
33-35 2 2
N=50
For example, we are looking for the 20th percentile or the score in which 20 percent of the cases
falls below it, then
20th percentile or P20 = LP20+ i[ (20%N – cf)/ f P20 ]
= 41.5 + 3 [(.20x50)-8]/6
= 41.5 + 3 [ (10-8)/6 ]
= 41.5 + 1.00
= 42.50
Note: The values of P20 and D2 are the same because in either measure, we are looking for the
score in which 20 percent of the cases falls below it.
Standard Deviation or SD
Standard Deviation or SD is a measure of dispersion among all scores in the distribution
rather than through extreme scores or only a proportion of the scores, making it the most reliable
measure of variability. Standard deviation is the square root of the average of the squared
deviation from the Mean.
Finding the Standard Deviation of Ungrouped Scores
Steps:
1. Find the Mean
2. Subtract the Mean from the scores
3. Square the deviation
4. Find the sum of the squared deviation (∑d2)
5. Divide the sum of the squared deviation by the number of cases
6. Find the square root of the answer in step 5
∑ d2
In symbol, SD=
Where:
√ N
d = deviation from the mean
d 2= squared deviation
∑ d 2 = sum of the squared deviation
Example 1. Compute for the Standard Deviations of the following scores:
92 75 85 83 90 73 79 80 88 83
Score d d2
92 +9 81
75 -8 64
85 +2 4
83 0 0
90 +7 49
73 -10 100
79 -4 16
80 -3 9
88 +5 25
83 0 0
∑d2= 348
N = 10
Mean = 83
∑ d2
SD=
√N
348
SD=
√
10
SD=√ 34.8
SD=5.59
SD=5.59
Finding the Standard Deviation of Grouped Scores
For ungrouped data, we compute for the standard deviation by computing first for the
actual Mean of the set of scores. For grouped data, standard deviation in computed from an
“assumed mean”. To illustrate, study the example given.
X f d fd fd2
75-77 3 5 15 75
72-74 4 4 16 64
69-71 6 3 18 54
66-68 5 2 10 20
63-65 8 1 8 +67 8
60-62 9 0 0 0
57-59 5 -1 -5 5
54-56 8 -2 -16 32
51-53 3 -3 -9 27
48-50 2 -4 -8 32
45-47 2 -5 -10 - 48 50
N=55 ∑fd = +19 ∑fd2 = 367
Steps:
1. Choose any step interval for the assumed mean as the arbitrary starting point or “origin”. In
the example given, the interval 60-62 has been chosen. Call this interval zero, and the next
higher interval +1, the lower interval -1, etc. These are shown in the column labeled d. (Note:
Any interval can be chosen, and the final result will be the same)
2. Multiply frequency (f) by the number of deviations (d) and the resulting product is shown in
column labeled fd. Get the sum of fd by taking into account the plus and minus signs.
3. To get fd2, multiply d by the fd. Then get the sum of fd2
Substitute the corresponding values to the formula:
SD=i
√ ∑ f d 2 − ∑ fd
N ( ) N
Where,
i= interval
N = number of cases
∑ fd = summation of frequency deviation
∑ f d 2= summation of frequency x squared deviation
2
367 19
SD=i
√ 55
− ( )
55
¿ 3 √ 6.67−( .119 )
¿ 3 √ 6.551
¿ 3∗2.55
SD=7.68
Summary
To represent the spread of scores, the range is the quickest measure. However, the range
is an unreliable measure because it greatly depends on extreme scores. Statisticians have
developed the semi-interquartile range, half the distance between the 25th and 75th percentile as a
more reliable measure of dispersion. We can also use deciles and percentiles to determine the
score by which certain percent of the cases fall below it. Finally, a measure of
variability/dispersion that involves all scores in the distribution is the standard deviation, a type
of average of the deviations of the scores away from the average (Mean). For ungrouped data,
standard deviation is computed from the actual meanwhile for the grouped data, it is computed
from an assumed mean of the distribution.
