Measurements and Experimental Results:

Graphical Analysis

Introduction:

Physics, like all science, relies on comparison between experimental observation

and scientific hypothesis. The “closeness” of this comparison tells us how
accurate and precise our hypothesis was in predicting the observed phenomenon.
Physicists usually employ mathematical models to describe their hypotheses
because math is a universal language and it allows all scientists across the world
to test hypotheses in as close similarity as possible.

The job of a physicist (or a physics student) is, fundamentally, threefold: 1) to

develop mathematical models that, to the best of our knowledge, will have the
ability to predict some observable phenomenon, 2) to develop a procedure that
produces consistently repeatable measurements associated with that observed
phenomenon, and 3) to determine and demonstrate a means of comparison
between the mathematical model and the measured data. The physicist then
returns to the first step and attempts to refine the mathematical model to be
more accurate and precise based on the results of the previous comparison.

This guide is intended to be a primer to help develop understanding for one of the
most basic means of comparison between mathematical models and measured
data, which we refer to as “graphical analysis”. This requires the unification of
algebra (and occasionally calculus) techniques with experimental science.

Review of the Pertinent Mathematics:

This is not a math textbook, but this section will help reinforce various
mathematical techniques that are integral to understanding the process, the
purpose, and the result of graphical analysis. The primary skills you will need for
the introductory physics lab courses are algebraic manipulation and plotting. To
review algebraic manipulation, please reference a textbook like College Algebra
(available at UCR’s Orbach Library).

Reference Acknowledgement: David Cook, Lawrence University

 2

You also need to understand how to relate mathematical constants to physical

parameters. For example, the equation for a straight line with non-zero y-
intercept looks like: y = A*x + B. In this equation, A and B are called parameters. In
this context, these parameters have no physical meaning, though they do have
mathematical meaning (A is “slope” and B is “y-intercept”). In physics, models are
developed so that these parameters do have physical meaning. For example, you
will come across the equation F = m*a, where F is force, m is mass, and a is
acceleration. This is a linear equation that looks like y = A*x + B where F = y, m =
A, x = a, and B = 0, so it is the same as the mathematical representation, only the
terms now have physical meaning, so they can be measured experimentally!

In the introductory physics lab courses, you will see several different physical
models, but all are characterized by just a few different mathematical models:
Proportional/Linear: y = A*x + B
Quadratic: y = A*x2 + B
Inverse: y = A/x + B = A*x-1 + B
Inverse Square: y = A/x2 + B = A*x-2 + B
Exponential: y = A*exp(B*x) + C
Sine: y = A*sin(B*x + C)
It will be important for you to be able to quickly visualize mathematical functions
in graphical form. Make sure you are familiar with each of these functions and
their graphical appearance. (Wolfram Alpha is a good reference for help.)

Mathematical Models to Describe Physical Hypotheses:

To develop mathematical models that Trend 1

Volume (m^3)

describe physical hypotheses, it is Trend 2

important to be able to visualize the Trend 3
model. For example, if we establish
the hypothesis: “A sphere’s volume
will increase when the radius is
increased,” we infer that the graphical
representation of that hypothesis Radius (m)
must look something like the figure at right. There are many different ways of
graphically representing this hypothesis. But which is correct?
 3

Well, for the hypothesis as it was stated, all three trends are correct, because the
hypothesis is not detailed enough to tell us what kind of relationship the trend
should have! This is where it is crucial to have physics concepts that underlie the
mathematical model. We can state a
y = A*x^3
more detailed hypothesis for the

Volume (m^3)
example above: the volume of a
sphere is proportional to the cube of
the radius, or mathematically: Volume
= constant*(radius)3. If we take this
mathematical form and plot it, we
produce the trend shown at right.
Radius (m)

Visualization of Measured Data:

Once you have a sufficiently detailed physical hypothesis and have visualized the
mathematical trend that that hypothesis should take, you are ready to collect
measurements to test that hypothesis! The visualization of the mathematical
trend must come before the measurements so that you can then quickly ascertain
qualitatively whether or not your measurement procedure has been properly
designed or if your hypothesis is clearly invalid.

When you make measurements in any scientific experiment, you are attempting
to observe a change in one variable dependent on some controlled change in
another variable. Returning to the cube volume example again, to test your
physical hypothesis, you might control a change in the side-length of the cube,
and you would then measure the resulting change in the cube’s volume. You
might come up with a data set and accompanying plot that looks like this:
0.6
Volume (m3) Radius (m) 0.5
Volume (m^3)

Measured Controlled 0.4

0.004 0.10 0.3
0.034 0.20 0.2
0.113 0.30
0.1
0.268 0.40
0
0.524 0.50 0 0.1 0.2 0.3 0.4 0.5 0.6
Radius (m)
 4

Note the plot convention – always plot your controlled variable on the horizontal
axis and your measured variable on the vertical axis. This makes comparisons
between mathematical models and measured data more straightforward. The
first step, after you have created this visualization of your measured data, is to
qualitatively compare your observations with your hypothesis. The hypothesis
states that the volume should increase as the side-length is increased, and that it
should be a non-linear increase. Both constraints are satisfied, so we can move on
to make a more quantitative comparison between measurement and model.

Relating Trends in Measured Data to Physical Hypotheses:

Once you have established a physical hypothesis and acquired some measured
data that qualitatively appears to validate that hypothesis, you are ready to move
on to the more quantitative comparison between measurement and prediction.
To do this, you will usually create a “best-fit line” using some analysis software.

To create the best-fit line, all you have to do is tell your software what kind of
mathematical model it should attempt to graph over your data. For the sphere-
volume example from above, this means combining the two plots on p. 3
together. Overlaying a mathematical model on top of some measured data is not
as trivial as it sounds. This is where the real importance of having well-established
parameters is absolutely necessary. 0.6
If you were just to plot a graph of y 0.5 Fit Function:
Volume (m^3)

3
= x along with your measured data, 0.4 y = x3
0.3
you would see something that looks
0.2
like the plot at right. There is clearly
0.1
no agreement between the model 0
(y = x3) and the measured volumes! 0 0.1 0.2 0.3 0.4 0.5 0.6
Radius (m)
This is why mathematical parameters are so important! Look at the table of
different mathematical models on p. 2. You will note that all models include some
constants like ‘A’, ‘B’, and ‘C’. These are the mathematical parameters associated
with the model. When you instruct software to generate a best-fit line according
to some designated mathematical model, it guesses some initial value of these
parameters and then iteratively changes those values until the model represents
the measured data as closely as possible. This is why it is called a “best-fit”.
 5

The y = x3 model didn’t fit our data 0.6

Volume (m^3)
very accurately because we didn’t Fit Function:
include any parameters that allow 0.4 y = A*x3
A = 4.19
the software to tune the best fit 0.2
curve. We need to include a
parameter that can be adjusted so 0
that the model is a good fit to the 0 0.1 0.2 0.3 0.4 0.5
data, so the model should actually Radius (m)
look like y = A*x3. The result is shown in the figure at right, and the fit is perfect!

The fitting software tells us what value of the parameter yields the closest fit to
the data; in this case A = 4.19. This is a mathematical value, not a physical value!
This parameter only has physical meaning if your hypothesis has established it! In
this example, we know there must be some value that is multiplied by the cube of
the radius, and since it is a sphere, we know that value must include a factor of π.
If we divide by π, we get A = 1.33*π, or A = 4/3*π. If we plug this back into our
initial model, it looks like Volume = 4/3*π*(radius)3, which we know to be valid!

A Final Note of Caution:

When doing experimental physics, you can never try to relate a mathematical
model to best-fit your measured data without having a physical hypothesis to
support the mathematical model! As an example, consider the measured data in
the plot below. This is real data that was collected and fit using two different
mathematical models. It is impossible to tell which model is “more correct”, and
both models have very different ramifications on the description of the physics!
ALWAYS make sure that you have a physical model underpinning any
mathematical trend you seek to fit, otherwise it is scientifically meaningless!
140 140
Damping Rate (1/s)

Damping Rate (1/s)

120 Fit with exponential model: 120 Fit with polynomial model:
100 y = A*exp(B*x) 100 y = A*x-3 + B
80 80
60 60
40 40
20 20
0 0
175 250 325 400 475 550 175 250 325 400 475 550
Magnetic Field (G) Magnetic Field (G)