This preview shows page 1. Sign up to view the full content.
Unformatted text preview: e can tell from Figure 4(b) that the object is at
at
. A grain
of salt: our prediction graphs will not always need units. For example, if we are asked to
draw a graph predicting the relationship of, say, the acceleration due to gravity of an object
with respect to its mass, the label “Mass” will do just fine for our horizontal axis. This is
because we are not expected to give the precise functional dependence in this situation, only
the overall behavior. We don’t know exactly what the acceleration will be at a mass of
,
and we don’t care. We just need to show whether the variation is increasing, decreasing,
constant, linear, quadratic, etc. In this specific case, it might be to our advantage to include
253 APPENDIX: REVIEW OF GRAPHS units on the vertical axis, though; we can probably predict a specific value of the acceleration,
and that value will be meaningless without them. (b) Position q with respect to time τ for a
mass of 3kg. The acceleration is constant. (a) Figure 4: Poorly versus welllabeled and captioned graphs. The labels and caption make the
second graph much easier to interpret.
Every graph we make should also have some sort of title or caption. This helps the reader
quickly to interpret the meaning of the graph without having to wonder what it’s trying to
say. It particularly helps in documents with lots of graphs. Typically, captions are more
useful than just titles. If we have some commentary about a graph, then it is appropriate to
put this in a caption, but not a title. Moreover, the first sentence in every caption should serve
the same role as a title: to tell the reader what information the graph is trying to show. In fact,
if we have an idea for the title of a graph, we can usually just put a period after it and let that
be the first “sentence” in a caption. For this reason, it is typically redundant to include both a
title and a caption. After the opening statement, the caption should add any information
important to the interpretation of a graph that the graph itself does not communicate; this
might be an approximation involved, an indication of the value of some quantity not
depicted in the graph, the functional form of a fit line, a statement about the errors, etc. Lastly,
it is also good explicitly to state any important conclusion that the graph is supposed to
support but does not obviously demonstrate. For example, let’s look at Figure 4 again. If we
are trying to demonstrate that the acceleration is constant, then we would not need to point
this out for a graph of the object’s acceleration with respect to time. Since we did not do that,
but apparently had some reason to plot position with respect to time instead, we wrote, “The
acceleration is constant.”
Lastly, we should choose the ranges of our axes so that our meaning is clear. Our axes do not 254 APPENDIX: REVIEW OF GRAPHS always need to include the origin; this may just make the graph more difficult to interpret.
Our data should typically occupy most of the graph to make it easier to interpret; see Figure 5.
However, if we are trying to demonstrate a functional form, some extra space beyond any
statistical error helps to prove our point; in Figure 5(c), the variation of the dependent with
respect to the independent variable is obscured by the random variation of the data. We must
be careful not to abuse the power that comes from freedom in (a) (b) Figure 5: Graphs with too much (a), just
enough (b), and too little space (c) to be
easy to interpret. (c)
plotting our data, however. Graphs can be and frequently are drawn in ways intended to
manipulate the perceptions of the audience, and this is a violation of scientific ethics. For
example, consider Figure 6. It appears that Candidate B has double the approval of
Candidate A, but a quick look at the vertical axis shows that the lead is actually less than one
part in seventy. The moral of the story is that our graphs should always be designed to
communicate our point, but not to create our point. 255 APPENDIX: REVIEW OF GRAPHS Figure 6: Approval
ratings
for
two
candidates in a mayoral
race. This graph is
designed to mislead the
reader into believing
that Candidate B has a
much higher approval
rating than Candidate
A. Using Linear Relationships to Make Graphs Clear
The easiest kind of graph to interpret is often a line. Our minds are very good at
interpreting lines. Unfortunately, data often follow nonlinear relationships, and our minds
are not nearly as good at interpreting those. It is sometimes to our advantage to force data to
be linear on our graph. There are two ways that we might want to do this in this class; one is
with calculus, and the other is by cleverly choosing what quantities to graph.
The “calculus” method is the simpler of the two. Don't let its name fool you: it doesn't
actually require any calculus. Let’s say that we want to compare the constant accelerations of
two objects, and we have data about their positions and velocities with respect to ti...
View
Full
Document
This document was uploaded on 02/23/2014 for the course MANAGMENT 2201 at University of Michigan.
 Spring '14

Click to edit the document details