IB Standard and Higher Level Physics

Error Analysis Rules and Methods

Significant Figures
Reporting significant digits is important for conveying the precision of measured quantities and maintaining consistency with
precision when carrying out mathematical operations. Remember: Your calculated answers cannot indicate higher precision than any
of the measurements that were used in the calculation. The following questions indicate how digits, when part of a physical
measurement, are either certain or uncertain. Every measurement has digits that the observer can state without a doubt is certain.
There will always be ONE single digit that has to be estimated. This is based on the precision of the measuring tool. The picture
below illustrates this.

The blue pencil is 42.23 cm. The nail is 41.00 cm. This measuring tool can estimate to the tenth of a millimeter, ALWAYS!

During the recording of measured values, how do you know how many digits to write down?

For example, I measure the length of a table to be 100.52 cm. In this measurement, I can be CERTAIN that the 1, 0, 0 and 5 digits are
correct. The last digit, 2, I had to estimate because it was in between the smallest marks (mm or 0.1 cm) on the meter stick.
Therefore, the last digit is UNCERTAIN.

RULE 0: For all measurements, make sure your last digit is the one that is UNCERTAIN.

The number 100.52 is said to have five significant figures. All digits that are non-zero are counted as significant. The zeroes are
special in that sometimes they are counted as significant, sometimes they are not.

RULE 1: Any zero that is IN BETWEEN two non-zero digits are counted as significant.
E.g. 100.25 12.5006 302.109
RULE 2: Any zero that is used for establishing the PRECISION OF A MEASUREMENT is counted as significant.
E.g. 13.500 125.30 365.03000
RULE 3: Any zero that is only used as a DECIMAL PLACEHOLDER is not counted as significant.
E.g. 520 135,000,000 0.00057
(Scientific notation fixes these ambiguities.)

How many significant figures are in each of the following numbers?

253 105.20 366.002 5730 0.000025
1.035006 23.5000 1002 905.3870 6400

How do you determine the number of digits to round your answers to when you add/subtract or multiply/divide?
RULE 4: When adding (or subtracting) numbers, your answer cannot be more precise than the least precise measurement.
e.g., 34212.5 + 3.256 = 34215.8
RULE 5: When multiplying (or dividing) numbers, your answer need only retain the same number of digits as the number that has the
least number of significant digits.
e.g., 25.36 * 2.2 = 56
How many significant figures are in each answer to the following problems?
36.25 + 3 9 – 0.653 32 x 2.35 56.75 / 253.65 65.897 + 3.25
89.36 - 5.2 1.035 x 3.6 12.48 / 3.2 1.9871 + 45.88 5.63 x 3.8925
Determining Uncertainty in Measurements
Random errors – Errors that arise out of measurement that cannot be controlled.
Systematic errors – Errors that arise due to imperfections in equipment or procedure. Sometimes called “calibration errors”
Precision – Shows us how good we are at repeating our measurements and getting consistent value, regardless of our accuracy. It is
an indicator of the number of significant figures your measurements can be reliably stated.
Absolute Uncertainty – Also known as Absolute error, it is the range of precision that your measurement exhibits, as an absolute
number.
Relative Uncertainty – Also known as Relative Error, it is the range of precision exhibited by your measurement, stated as a
percentage of your absolute uncertainty out of the absolute measurement.
Accuracy – Shows us how close our measurement gets to the accepted value, regardless of our precision.
Percent Error – The percent by which your experimentally measured results differs from the theoretically accepted value.

With these two concepts in mind, it is important that every measurement you take has a stated UNCERTAINTY that indicates the
most likely range of possible values that your measurement was rounded from.

E.G. My weight = 201 ± 1 lbs. This means that when reading the scale, I am confident that my true weight is somewhere in
between 200 and 202 lbs. This would be an ACCURATE measurement of my weight. However, if I stated my weight as
201 ± 15 lbs, this would not be as accurate. Likewise, if I had stated my weight as 200.35 ± 0.05 lbs, then I am confident that
my weight is both ACCURATE and PRECISE. How might I state my weight as a precise measurement but NOT accurate?
Note that the precision in the uncertainty must match the precision in the measured value.

RULE 6: Uncertainties in Single Measurements

An uncertainty can be established in a single measurement by estimation alone, such as measuring using a meter stick. Measurements
such as these are easily repeatable. This method is not always the best choice if there is a high degree of random error, such as using a
stopwatch for recording the time of an event that is difficult to repeat.

RULE 7: Repeatable Measurements Uncertainty Estimates

An uncertainty can be established for a single measureable quantity by repeating it several times and then estimating it by looking at
the range of high and low values. The average of the trials can be taken as the measurement of the quantity and the uncertainty is
based on the range of values in the trials.

E.g. The fall times for dropping an object are recorded as 5 trials. Their times are 2.56 s, 2.58 s, 2.55 s, 2.59 s and 2.53 s.
Determine the average time of the falling object along with its uncertainty. Remember to use the correct number of
significant digits.

An important note: Estimating the uncertainty in this way usually OVERESTIMATES it because you are basing it on only a
limited number of trials. A better method (though not required of you) would be to use the SAMPLE STANDARD
DEVIATION as the best estimate of uncertainty. This is explained on the following pages.
RULE 8: Standard Deviation as the Uncertainty in Repeated Measurements
 Sometimes the confidence of a measurement cannot be reliable using one data point. A time measurement might be an
example, especially if one is using a stopwatch. Repeated measurements can be used to account for random errors due to
uncertainties in human reaction time by taking an average value. The standard deviation is a statistical quantity that tells us
how the repeated measurements are spread out. In general, data that is subject to random errors tends to fall under a
distribution curve known as a Gaussian or Normal.

If a large number of single measurements were taken and they followed the above patterns (e.g. the weight of 5 year old baby African
elephants for the bottom one and average human IQ on the Simon-Binet Scale for the top one, then a standard deviation above and
below the average value would indicate that 68% of the test subjects occupy this region of the graph. For uncertainties, if we said that
the a 5 year old baby African elephant weighs 500 lbs, we could say that it weighs 500 ± 100 lbs, telling us that we are 68% sure that
if we picked a random 5 year old baby African elephant, it would weigh between 400 and 700 pounds.

We use an uncertainty to estimate our confidence in our measured value, especially if we are comparing it to a theoretical value, such
as gravity.

The standard deviation formula is (the N-1 factor calculates a “Sample Standard Deviation” which assumes that you are sampling a
smaller portion of a larger pool of potential data. This is more than likely the case in statistical science.)

s = standard deviation N = number of measurements

x bar = average of N measurements xi = the ith measurement

Standard Deviation Practice

Find the uncertainty in the following set of numbers assuming that they are measurements of a single variable taken as a sample set.
45, 67, 54, 34, 26, 59, 23, 69, 53, 39
Determining Uncertainties in Calculations

Since individual measurements have a degree of error/uncertainty, it stands to reason that any operations that involve making
calculations with these measurements would act on the errors as well, making them compound the error in your calculated result.
Mathematical operations, such as basic equations in physics, will produce answers that have larger errors/uncertainties than the
original measurements. There are specific rules for PROPAGATING these errors and are outlined below.

1. Propagating uncertainties in a sum of numbers.

If we know that A = B + C, then the uncertainty in A, which we call δA, is given by the formula below, where δB and δC are the
uncertainties in the measured values B and C. Note the Greek symbol  . This is the lowercase symbol for “delta”, which is
customarily used to represent “uncertainty”. These represent Absolute Uncertainty.

A  B  C
2. Propagating uncertainties in an equation with a coefficient.

If we know that A = cB, then the uncertainty in A, which we call δA, is given by the formula below, where δB is the uncertainty
in the measured value of B.
A  cB
3. Propagating uncertainties in a product or quotient of numbers.

If we know that A = BC/D, then the percent (or fractional) uncertainty in A, which we call δA/A, is given by the formula below,
where δB/B, δC/C and δD/D are the percent uncertainties in the measured quantities B, C and D. These represent Relative
Uncertainty.
A B C D
  
A B C D
Equivalently: Sometimes it is easier to propagate uncertainties in products or quotients by explicitly finding the percent error.
With sums and differences, we state that the uncertainty of the sum or difference of two or more numbers is the sum of the
uncertainties of the numbers that are being added or subtracted, as discussed in (1) above. For products and/or quotients, we can
state that the PERCENT UNCERTAINTY of the product/quotient of two or more numbers is the SUM of the PERCENT
UNCERTAINTIES of all the numbers that you are multiplying or dividing. From percent uncertainties we can determine
absolute uncertainties. These represent Relative Uncertainty.
% A  % B  % C  % D

4. Propagating uncertainties in a power.

If A = Bc, then the uncertainty in A is given by the formula below.
A B
 c
A B

5. Any function of one variable.

If a variable x, is measured to have an uncertainty δx and it is used to evaluate a function q(x), then the uncertainty in the
calculated value of q(x), δq is given by the following formula: You should not worry about this possibility if you have not
explored differential calculus yet. The physics that you will explore will only deal with the possibilities listed above. The
equation below could prove useful if you choose to do an IB Extended Essay in Physics as part of your Diploma Program.
dq
q  x
dx
Practice with Uncertainties

Propagating uncertainties in a sum of numbers. Solve the following using the correct number of significant digits and determine
the uncertainty in the following sums:

1. (1.05 ± 0.02) + (3.27 ± 0.03) = _______________________

2. (35.0 ± 5.0) + (4.8 ± 0.1) = _______________________

3. (0.045 ± 0.013) - (0.029 ± 0.090) = _______________________

4. (4.1 ± 1.0) - (2.19 ± 0.14) = _______________________

5. (101.050 ± 0.017) + (53.3 ± 0.1) = _______________________

Propagating uncertainties in a product of numbers. Solve the following using the correct number of significant digits and
determine the uncertainty in the following products: (Yes the numbers are the same, only the operation has changed!)

6. (1.05 ± 0.02) x (3.27 ± 0.03) = _______________________

7. (35.0 ± 5.0) x (4.8 ± 0.1) = _______________________

8. (0.045 ± 0.013) / (0.029 ± 0.09) = _______________________

9. (4.1 ± 1.0) / (2.19 ± 0.14) = _______________________

10. (101.050 ± 0.017) x (53.3 ± 0.1) = _______________________

Propagating uncertainties in an equation with coefficients. Take the following equation:

y = 3x

For the following values of x, determine the value of y along with its uncertainty:

11. x = 54 ± 4

12. x = 45.7 ± 0.5

Propagating uncertainties with power equations. Take the following equation:

y = x3
For the following values of x, determine the value of y along

13. x = 8 ± 3

14. x = 0.036 ± 0.010

Uncertainty and Error Analysis in Graphing:
Suppose you have a graph of data points and all the uncertainties in your measurements are depicted on the graph using error bars.
There is an implication that uncertainties in measurements that are plotted on a graph will lead to uncertainties in all calculations
involving the graph. As an example, a line of best fit on a graph of plotted points, say for velocity vs. time, should have calculable
uncertainties in its slope and uncertainties. How would you go about doing this using a strictly graphical technique and not via
statistical methods? IB requires that you know how to do this graphically. At a later date, we can discuss the possibility of analyzing
graphs using statistical methods but it is not a requirement.

Here is a sample graph of velocity vs. time data with error bars.

m = 2.404 ± 0.677 ms-2

Y-Int = -0.334 ± 2.721 ms-1

The slope of this line is 2.404 ms-2 and the y-intercept is -0.334 ms-1. These are calculated from the original line of best fit. Work
should be shown. How would you determine the uncertainties in the slope and the y-intercept?

The error bars tell all you need to know. Above and below the central line of best fit are grayed lines that pass through the ends of the
error bars on select points near the beginning and end of the plotted data points. The slopes of the grayed lines indicate lines of best fit
for maximum and minimum slope possibilities, based on uncertainties in measuring the data. The process below shows how to
calculate the uncertainty in the slope.
Slope Uncertainty:
 The slope of the steeper gray line is 3.095 ms-2 which is 0.691 ms-2 greater than the actual slope.
 The slope of the less steep line is 1.741 ms-2 which is 0.663 ms-2 less than the actual slope.
 The average value of these differences is 0.677 ms-2.

The slope is then stated as 2.404 ± 0.677 ms-2.

Y-Intercept Uncertainty:
 Similar arguments are presented for the y-intercept and its published uncertainty.
 From the graph the y-intercept for the steeper sloped line is 2.094 ms-1.
 For the less steep line it is -3.347 ms-1.
 The larger y-intercept is 2.428 ms-1 greater than the actual y-intercept.
 For the less steep line it is 3.013 ms-1 less than the actual y-intercept.
 The average of these differences is 2.721 ms-1.

The y-intercept is then stated as -0.334 ± 2.721 ms-1.

You would need to show these calculations and write in your results on the graph, as indicated.

This method is standard for extracting information from graphs of plotted experimental data. For example, the graph shown above is
for velocity vs. time. The slope of this graph is a representation of ACCELERATION. By measuring the slope and determining its
uncertainty we have calculated the acceleration of an object rolling down an incline plane along with a statement of the uncertainty in
our calculation, merely using graphical techniques.
Graphical Analysis Practice
Plot the following data on the graph below, Mass on the vertical axis and Volume on the horizontal axis. Determine both the slope
and y-intercepts along with their uncertainties as described previously in this packet. Include units for the slope and uncertainties.
Mass
2.0 4.0 6.0 8.0 10.0
(± 1.0 g)
Volume
8.7 21.2 29.8 43.1 49.9
(± 0.5 cm3)
Calculations for Graphical Analysis
 Comments on Precision, Accuracy and Percent Error
The information obtained from a graph, as shown above, is often used for subsequent calculations. For
example, given the experimental value for the acceleration, suppose you wanted to calculate an experimental
value for gravity and you knew that the acceleration of an object on an incline plane is related to gravity by the
2𝑠
following equation 𝑔 = . In order to determine the acceleration of gravity, you would need to know the
𝑎𝑠𝑖𝑛𝜃
distance the rolling object traveled down the incline plane, s and the angle of the incline plane, θ. The
acceleration, a, is the slope you just calculated. The distance and angle are easily measured from experiment
but they also have associated uncertainties. Doing the multiplication will give you an answer for the
acceleration of gravity but you must also state the uncertainty in this calculation. You will be provided ample
practice in propagating uncertainties and graphical analysis at the end of this packet.

A good experiment, based on accuracy and precision, will have the true value of gravity’s acceleration fall
within the limits of uncertainty of your experimental value.

Accurate and Precise Scenario

For example, suppose your experiment revealed that g = 9.89 ± 0.20 ms-2. Since the true value of 9.80 ms-2
falls within these boundaries, you can confidently say that your measurement techniques were both
ACCURATE and PRECISE.

Accurate but Not Precise Scenario

Alternatively, suppose your experiment showed that g = 9.85 ± 1.50 ms-2. Although the true value of 9.80 ms-2
still falls with these boundaries, the uncertainties are abnormally large. In this case we have an experiment that
proved to be ACCURATE but NOT PRECISE.

Not Accurate but Precise Scenario

Finally, there is the case where we might have an experimental value of gravity, g = 10.85 ± 0.03 ms-2. In this
case the measurement is extraordinarily precise, but the true value falls outside these boundaries. This is an
experiment that is NOT ACCURATE but PRECISE.
Bizarre but Possible Scenario
There is another remote scenario that needs to be discussed. Consider a situation where you had extraordinary
precision and accuracy. As an example, you calculated gravity’s acceleration to be 9.83 ± 0.01 ms-2. Even
though the value 9.83 is very close to 9.80, we would have to say the experiment was both ACCURATE and
PRECISE, but the true value still lies outside the boundaries of the uncertainties. What does this mean?
Ironically, our measuring tools do show good accuracy and precision but with an uncertainty of only 0.01 ms-2.
It means we could have done better than experimentally arriving at a value of 9.83 ms-2. This is a possible
indication of a calibration error in our equipment.

Percent Error Calculations

The percent error of an experimentally determined value compared to a theoretical variable gives you a gauge
for determining the accuracy of your measurements. The formula for percent error is given below.
|𝑬𝒙𝒑𝒆𝒓𝒊𝒎𝒆𝒏𝒕𝒂𝒍 − 𝑻𝒉𝒆𝒐𝒓𝒆𝒕𝒊𝒄𝒂𝒍|
%𝑬𝒓𝒓𝒐𝒓 = 𝒙𝟏𝟎𝟎%
𝑻𝒉𝒆𝒐𝒓𝒆𝒕𝒊𝒄𝒂𝒍

A negative percent error merely means that our experimental error was less than the theoretical value.

For the four scenarios described above, determine the %Error of the experimental values of gravity compared to
the theoretical. For a high school physics lab, I consider less than 5% an indicator of an accurate lab. In
university, I was told to shoot for less than 3%. Ha Ha! LOL!!