Examples of Uncertainty calculations

• Uncertainty in a single measurement

• Fractional and percentage uncertainty
• Combining uncertainties in several quantities: adding or subtracting
• Combining uncertainties in several quantities: multiplying or dividing
• Is one result consistent with another?
• What if there are several measurements of the same quantity?
• How can one estimate the uncertainty of a slope on a graph?

Uncertainty in a single measurement

Bob weighs himself on his bathroom scale. The smallest divisions on the scale are 1-pound
marks, so the least count of the instrument is 1 pound.
Bob reads his weight as closest to the 142-pound mark. He knows his weight must be larger than
141.5 pounds (or else it would be closer to the 141-pound mark), but smaller than 142.5 pounds
(or else it would be closer to the 143-pound mark). So Bob's weight must be
weight = 142 +/- 0.5 pounds
In general, the uncertainty in a single measurement from a single instrument is half the least
count of the instrument.

Fractional and percentage uncertainty

What is the fractional uncertainty in Bob's weight?
uncertainty in weight
fractional uncertainty = ------------------------
value for weight

0.5 pounds
= ------------- = 0.0035
142 pounds

What is the uncertainty in Bob's weight, expressed as a percentage of his weight?

uncertainty in weight
percentage uncertainty = ----------------------- * 100%
value for weight

0.5 pounds
= ------------ * 100% = 0.35%
142 pounds

Combining uncertainties in several quantities: adding or subtracting

When one adds or subtracts several measurements together, one simply adds together the
uncertainties to find the uncertainty in the sum.
Dick and Jane are acrobats. Dick is 186 +/- 2 cm tall, and Jane is 147 +/- 3 cm tall. If Jane stands
on top of Dick's head, how far is her head above the ground?
combined height = 186 cm + 147 cm

= 333 cm

uncertainty in combined height = 2 cm + 3 cm

= 5 cm

combined height = 333 cm +/- 5 cm

Now, if all the quantities have roughly the same magnitude and uncertainty -- as in the example
above -- the result makes perfect sense. But if one tries to add together very different quantities,
one ends up with a funny-looking uncertainty. For example, suppose that Dick balances on his
head a flea (ick!) instead of Jane. Using a pair of calipers, Dick measures the flea to have a
height of 0.020 cm +/- 0.003 cm. If we follow the rules, we find
combined height = 186 cm + 0.020 cm

= 186.020 cm

uncertainty in combined height = 2 cm + 0.003 cm

= 2.003 cm

??? combined height = 186.020 cm +/- 2.003 cm ???

But wait a minute! This doesn't make any sense! If we can't tell exactly where the top of Dick's
head is to within a couple of cm, what difference does it make if the flea is 0.020 cm or 0.021 cm
tall? In technical terms, the number of significant figures required to express the sum of the two
heights is far more than either measurement justifies. In plain English, the uncertainty in Dick's
height swamps the uncertainty in the flea's height; in fact, it swamps the flea's own height
completely. A good scientist would say
combined height = 186 cm +/- 2 cm
because anything else is unjustified.

Combining uncertainties in several quantities: multiplying and dividing

When one multiplies or divides several measurements together, one can often determine the
fractional (or percentage) uncertainty in the final result simply by adding the uncertainties in the
several quantities.
Jane needs to calculate the volume of her pool, so that she knows how much water she'll need to
fill it. She measures the length, width, and height:

length L = 5.56 +/- 0.14 meters

= 5.56 m +/- 2.5%
 width W = 3.12 +/- 0.08 meters
= 3.12 m +/- 2.6%

depth D = 2.94 +/- 0.11 meters

= 2.94 m +/- 3.7%

To calculate the volume, she multiplies together the length, width and depth:
volume = L * W * D = (5.56 m) * (3.12 m) * (2.94 m)

= 51.00 m^3
In this situation, since each measurement enters the calculation as a multiple to the first power
(not squared or cubed), one can find the percentage uncertainty in the result by adding
together the percentage uncertainties in each individual measurement:

percentage uncertainty in volume = (percentage uncertainty in L) +

(percentage uncertainty in W) +
(percentage uncertainty in D)

= 2.5% + 2.6% + 3.7%

= 8.8%
Therefore, the uncertainty in the volume (expressed in cubic meters, rather than a percentage) is

uncertainty in volume = (volume) * (percentage uncertainty in volume)

= (51.00 m^3) * (8.8%)

= 4.49 m^3
Therefore,
volume = 51.00 +/- 4.49 m^3
= 51.00 m +/- 8.8%
If one quantity appears in a calculation raised to a power p, it's the same as multiplying the
quantity p times; one can use the same rule, like so:
Fred's pool is a perfect cube. He measures the length of one side to be
length L = 8.03 +/- 0.25 meters
= 8.03 m +/- 3.1%
The volume of Fred's cubical pool is simply
3
volume = L

volume = L * L * L = (8.03 m) * (8.03 m) * (8.03 m)

= 517.8 m^3
Just as before, one can calculate the uncertainty in the volume by adding the percentage
uncertainties in each quantity:
percentage uncertainty in volume = (percentage uncertainty in L) +
(percentage uncertainty in L) +
(percentage uncertainty in L)
 = 3.1% + 3.1% + 3.1%

= 9.3%
But another way to write this is using the power p = 3 times the uncertainty in the length:

percentage uncertainty in volume = 3 * (percentage uncertainty in L)

= 3 * 3.1%

= 9.3%
When the power is not an integer, you must use this technique of multiplying the percentage
uncertainty in a quantity by the power to which it is raised. If the power is negative, discard the
negative sign for uncertainty calculations only.

Is one result consistent with another?

Jane's measurements of her pool's volume yield the result
volume = 51.00 +/- 4.49 m^3
When she asks her neighbor to guess the volume, he replies "54 cubic meters." Are the two
estimates consistent with each other?
In order for two values to be consistent within the uncertainties, one should lie within the range
of the other. Jane's measurements yield a range
51.00 - 4.49 m^3 < volume < 51.00 + 4.49 m^3

46.51 m^3 < volume < 55.49 m^3

The neighbor's value of 54 cubic meters lies within this range, so Jane's estimate and her
neighbor's are consistent within the estimated uncertainty.

What if there are several measurements of the same quantity?

Joe is making banana cream pie. The recipe calls for exactly 16 ounces of mashed banana. Joe
mashes three bananas, then puts the bowl of pulp onto a scale. After subtracting the weight of the
bowl, he finds a value of 15.5 ounces.
Not satisified with this answer, he makes several more measurements, removing the bowl from
the scale and replacing it between each measurement. Strangely enough, the values he reads from
the scale are slightly different each time:
15.5, 16.4, 16.1, 15.9, 16.6 ounces
Joe can calculate the average weight of the bananas:
15.5 + 16.4 + 16.1 + 15.9 + 16.6 ounces
average = -------------------------------------------
5

= 80.4 ounces / 5 = 16.08 ounces

Now, Joe wants to know just how flaky his scale is. There are two ways he can describe the
scatter in his measurements.
 • The mean deviation from the mean is the sum of the absolute values of the differences
between each measurement and the average, divided by the number of measurements:
• 0.5 + 0.4 + 0.1 + 0.1 + 0.6 ounces
• mean dev from mean = --------------------------------------
• 5
•
• = 1.6 ounces / 5 = 0.32 ounces
• The standard deviation from the mean is the square root of the sum of the squares of
the differences between each measurement and the average, divided by one less than the
number of measurements:
• [ (0.5)^2 + (0.4)^2 + (0.1)^2 + (0.1)^2 +
0.6)^2 ]
• stdev from mean = sqrt
[ ----------------------------------------------- ]
• [ 5 - 1
]
•
• [ 0.79 ounces^2 ]
• = sqrt [ -------------- ]
• [ 4 ]
•
• = 0.44 ounces
Either the mean deviation from the mean, or the standard deviation from the mean, gives a
reasonable description of the scatter of data around its mean value.
Can Joe use his mashed banana to make the pie? Well, based on his measurements, he estimates
that the true weight of his bowlful is (using mean deviation from the mean)
16.08 - 0.32 ounces < true weight < 16.08 + 0.32 ounces

15.76 ounces < true weight < 16.40 ounces

The recipe's requirement of 16.0 ounces falls within this range, so Joe is justified in using his
bowlful to make the recipe.

How can one estimate the uncertainty of a slope on a graph?

If one has more than a few points on a graph, one should calculate the uncertainty in the slope as
follows. In the picture below, the data points are shown by small, filled, black circles; each
datum has error bars to indicate the uncertainty in each measurement. It appears that current is
measured to +/- 2.5 milliamps, and voltage to about +/- 0.1 volts. The hollow triangles represent
points used to calculate slopes. Notice how I picked points near the ends of the lines to calculate
the slopes!
 1. Draw the "best" line through all the points, taking into account the error bars. Measure
the slope of this line.
2. Draw the "min" line -- the one with as small a slope as you think reasonable (taking into
account error bars), while still doing a fair job of representing all the data. Measure the
slope of this line.
3. Draw the "max" line -- the one with as large a slope as you think reasonable (taking into
account error bars), while still doing a fair job of representing all the data. Measure the
slope of this line.
4. Calculate the uncertainty in the slope as one-half of the difference between max and min
slopes.
In the example above, I find
147 mA - 107 mA mA
"best" slope = ------------------ = 7.27 ----
10 V - 4.5 V V

145 mA - 115 mA mA
 "min" slope = ------------------ = 5.45 ----
10.5 V - 5.0 V V

152 mA - 106 mA mA
"max" slope = ------------------ = 9.20 ----
10 V - 5.0 V V

mA
Uncertainty in slope is 0.5 * (9.20 - 5.45) = 1.875 ----
V
There are at most two significant digits in the slope, based on the uncertainty. So, I would say the
graph shows
mA
slope = 7.3 +/- 1.9 ----
V

Percent difference
From Wikipedia, the free encyclopedia
Jump to: navigation, search

Percent difference or relative percent difference (RPD) is the numerical interpretation of

comparing two values with one another. It is often used as a quantitative indicator of quality
assurance and quality control for repeated measurements where the outcome is expected to be the
same.
The general requirement for selecting two values to be compared is that the user of this technique
expects the two values to be numerically equivalent. In other words, obtaining a percent
difference of 0% is the optimum result as it explains that the two values are exactly the same.
Not a general requirement, but common use shows that the two values usually will pertain to the
same property of an object (let's say the mass of an object, a material's characteristic, or maybe
the discharging time of a capacitor), but each value will be calculated using two different
methods and/or theories. Emphasis must be made on the word calculated, because the most
important requirement for the two values that are being compared using percent difference is that
they had needed to be calculated indirectly of measurement of the objective value. In other
words, neither of the two values can be the actual or accepted value of the objective value.
The two values are determined using theories and measurements that are to be tested with respect
to an accepted value by the scientific community. Example is someone measuring the length and
period of a pendulum to determine the acceleration of gravity, but must relate the value to the
accepted value of gravity by the scientific community.
Percent difference is similar to another comparison technique called percent error (note that a
more informative result comes from a non-absolute function, this will be discussed later), which
is when one determines an experimental value and is comparing it to the accepted or actual
value. Percent difference is different in that neither of the two values are the accepted or actual
value; they are both an experimental value determined by two different techniques but describing
the same objective value.

Contents
[hide]
 • 1 Formula
• 2 Percent error
• 3 References
• 4 See also

[edit] Formula
The equation for determining the percent difference, (assuming both x1 and x2 are positive), by
comparing values x1 and x2 is:

In sentence form, one is dividing the absolute difference of the two values by the average value
of x1 and x2. Because this equation contains the absolute function, percent difference will always
be positive (except when x1 and x2 are both negative) and therefore it does not matter which
value one assigns to the variables (x1 and x2) used in these equations shown in this article. A
simplified form of the equation can be calculated as,

Both values (x1 and x2) must contain the same units in order to be compared correctly with one
another. And as mentioned before, a zero percent difference is optimum and the higher the
percent value, the less precision of the two values.
One final note to make is that a lot of confusion lies in mistakenly assuming that percent
difference is the same as percent error. The difference is that percent difference is comparing two
experimental values, whereas percent error compares one experimental value with the
actual/accepted value.

[edit] Percent error

It seems the general standard of calculating the percent error involves using the absolute function
imposed on the difference between the experimental (measured) and accepted (actual) values.
However, this removes detail from the result in the form of only producing a positive percent
error value. It should be suggested to ignore the absolute function and calculate the percent error
as follows,

The numerator should be the Experimental value minus the Accepted value and not the other
way around. By using the equation shown above, the result will be positive only when the
experimental value is greater than the accepted and the result will be negative only when the
experimental value is less than the accepted.
This is a very important outcome. By avoiding the absolute function when calculating for the
percent error, the results will give both the reader and author more information. If the percent
error is negative, the reader knows immediately that the experimental value is short of the
accepted (goal) value. If the percent error is positive, the reader knows that the experimental
value is above the accepted (goal) value. This technique of solving the percent error value
becomes very helpful whenever an accepted value imposes a lower or upper limit for all
experimental (measured) values.
A rough example would be the goal to determine the speed of light. If an experiment produced a
speed that is greater than the speed of light, the reader will know immediately from a positive
percent error that something is wrong. The actual value will place the "no greater than" limit on
all measured values. Therefore, only negative percent errors should be expected! Using the
absolute function will hide this insight and important information and could be devastating to
relevant experiments.
Physics Lab Tutorials

Often, laboratory experiments are designed to determine the value of well-known physical quantities such
as the value of π or the acceleration due to earth's gravity, g. Since these quantities have accepted or
true values, we can calculate the percent error between our measurement of the value and the
accepted value with the formula

Sometimes, we will compare the results of two measurements of the same quantity. For instance, we may
use two different methods to determine the speed of a rolling body. In this case, since there is not one
accepted value for the speed of a rolling body, we will use the percent difference to determine the
similarity of the measurements. This is found by dividing the absolute difference of the two measured
values by their average, or