Professional Documents
Culture Documents
be some difference between the true value and the measured value. The
terms accuracy and precision are used to describe different aspects of the
difference between them. From a scientific point of view, these have very
different meanings.
The accuracy of a measurement is determined by how close the result of the
measurement is to the true value. For example, several experiments
determine a value for the acceleration due to gravity. For this case the
accuracy of the result is decided by how close it is to the true value of 9.80
m/s2. For several of the laboratory experiments, though, the true value of the
measured quantity is not known and the accuracy of the experiment cannot
be determined from the available data.
The precision of a measurement refers essentially to how many significant
digits there are in the result. It is an indication also of how reproducible the
results are when measurements of some quantity are repeated. When
repeated measurements of some quantity are made, the mean of those
measurements is considered to be the best estimate of the true value. The
smaller the variation of the individual measurements from the mean, the
more precise the quoted value of the mean is considered to be. This idea
about the relationship between the size of the variations from the mean and
the precision of the measurement shall be elaborated later.
Lets look at the following table of values acceleration due to gravity
measured by four students:
Measurement
1
Measurement
2
Measurement
3
Mean
Alf
7.83
dev Beth
1.60 9.53
dev Carl
0.27 8.70
dev Dee
0.04 9.72
dev
0.04
11.61
2.18
9.38
0.12 8.75
0.01 9.86
0.10
8.85
0.58 8.87
0.39 8.77
0.03 9.70
0.06
9.43
9.26
8.74
9.76
Notice by the definitions of accuracy and precision that Dees value of 9.76
is the most accurate while Carls is the least accurate, and Carls value is the
most precise while Alfs is the least precise.
Notice the interplay between the concepts of accuracy and precision that
must be considered. If a measurement appears to be very accurate, but the
precision is poor, the question arises whether or not the results are really
meaningful. Consider Alfs mean of 9.43, which differs from the true value
of 9.80 by only 0.37 and thus appears to be quite accurate. However, all of
his measurements have deviations greater than 0.37, and two of his
deviations are much larger than 0.37. It seems much more likely than that
Alfs mean of 9.43 is due to luck than to a careful measurement. If seems
likely, however, that Dees mean of 9.76 is meaningful because the
deviations of her individual measurements from the mean are small. In other
words, unless a measurement has high precision it cannot really be
considered to be accurate.
An examination of the significant figures given in this data leads to
essentially the same evaluation of each students data. Consider Alfs data,
which indicates by the values stated for the individual measurements that
two places to the right the decimal point are significant. However, that
conclusion is not supported by the fact that his deviations occur in the first
digit to the left of the decimal point. On the other hand, Dees results show
deviations in the second place to the right of the decimal point in agreement
with the fact that two places to the right of the decimal are given as
significant in the measured values. Thus from another point of view, Dees
results are seen as meaningful, but Alfs are questionable.
Carls results, on the other hand, are an example of a situation that is
common the interplay between accuracy and precision. Carls precision is
extremely high yet his accuracy is not very good. When a measurement has
high precision but poor accuracy, it is often the sign of a systematic error. A
systematic error is an error that tends to be in the same direction for repeated
measurements, giving results that are either consistently above the true value
or consistently below the true value. In many cases such errors are caused
by some flaw in the experimental apparatus, like not calibrating a device
correctly. Another source of a systematic error is failing to take into account
all of the variables that are important in the experiment. For Carl, if all his
value were consistently below the true value, this might represent Carl
forgetting to take into account friction, which would indeed cause all his
values to be low. But since his mean is well above the true value, this points
to a systematic error involving the equipment.
Percent Error and Difference
In several laboratories, the true value of the quantity being measured will be
known. In those cases, the accuracy of the experiment will be determined by
comparing the best estimate of the true value, or experimental value, with
the known true value. This can be done by figuring the percentage deviation
from the known true value (as known as percentage error). If E stands for
experimental value, and K stands for the known value, then:
Percentage error
EK
K
x 100%
E2 E1
Percentage difference
x 100%
E2 E1
centimeter. You measure the length and find that it falls about halfway 27
and 28 cm. You estimate that the length is 27.5 cm, but the 0.5 cm is not
exact, but a guess, so you could report that the length of the string to be
27.5 0.1 cm. When a number is reported, typically the number of digits
reported is the number known with any certainty. The uncertainty is
generally assumed to be one or two units of the last digit, but may be
different depending on the situation.
For example,
Number
1.2
3.61
19.61
0.017
9.504
0.1020
Number of Significant
Figures
2
3
4
2
4
4
753.1
37.08
0.697
56.3__
847.177 847.2
327.23
x 36.73
12019.158 12020
8.90906 8.909
36.73 327.23
Propagation of Uncertainties
Assume we have a function of two variables x and y, f(x,y). The variables x
and y are assumed to be independent in the sense that each can vary
arbitrarily without affecting the other. To find the change in f due to small
changes in x and y, we calculate the total differential of f. For example, if f
is a function for determining area, f = xy, then,
df = xdy + ydx
where the bars denote absolute value, and where we have changed our
notation to reflect the fact that our uncertainties are not true differentials,
i.e., we have made the identifications
df f ,
dx x,
dy y
f ( x, y ) x y
f ( x, y ) x
y
f ( x) x k
Uncertainty formula
f x y
f
x
y
f
x
y
f
x
k
f
x
Note that this includes the commonly occurring cases of x2 and x0.5
f ( x ) sin x
f ( x ) cos x
f (cos x ) x
f (sin x ) x
For example, assume four measurements are made of some quantity x, and
the four results are 18.6, 19.3, 17.7, and 20.4. By the above equation, the
mean value is:
1
18.6 19.3 17.7 20.4 19.0
4
n 1
x
n
1
2
2
2
2
(18.6 19.0) (19.3 19.0) (17.7 19.0) ( 20.4 19.0) 1.1
4 1
where y is the slope of the line and b is the y-intercept (the value of y at x =
0). Often, though, because of random errors, a graph of the data does not
display a perfectly linear relationship where every data point lies exactly on
a straight line. So it would be convenient to determine the value of m and b
that produces the best straight line fit to the data. Any choice of values for
m and b will produce a straight line, with values of y determined by the
choice of x. For any such straight line, there will be a deviation between
each of the measured ys from the data points and the ys from the straight
line fit at the value of the measured xs from the data points. The linear least
squares fit is that m and b for which the sum of the squares of these
deviations is a minimum. The linear least squares fit process is also called
just linear fit or linear regression.
There is a quantitative measure of how well the data follow the straight line
obtained by the least squares fit. It is given by the value of a quantity called
the correlation coefficient (sometimes referenced as r or R2). This quantity
is a measure of the fit of the data to a straight line with R2 = 1.00 signifying
Velocity (m/s)
3.0
3.7
4.4
6.0
Velocity (m/s)
5.5
5
4.5
4
v = 0.7514t + 2.2086
3.5
R = 0.9986
3
2.5
2
0
Time (sec)
Laboratory Reports
The following paragraphs will give you an idea of what a report should
contain and what it should accomplish. The order in which the contents
appear is of secondary importance, but they must all be there, clearly stated
and in a manner suited to the particular situation.
The purpose of a lab report such as this is to communicate fully and
coherently the results of an experimental investigation. Persons outside the
immediate group performing the experimental work may be interested in the
results of your work. It is therefore important that these results and your
analysis of them be put forth in a form suitable for such communication.
The effectiveness of this communication will depend on the clarity and
conciseness of your explanation of the work and the facility with which
desired material can be extracted from it.
Although there is no set universal form for reporting laboratory work, due to
the varied nature of experimentation, there are a few basic things which all
technical reports on experimental work should do. A lab report should at the
very beginning tell the reader what the experiment was for; what it
attempted to find out. Secondly, the report should inform the reader, in a
concise way, how the work performed is expected to accomplish this
purpose. Thirdly, where warranted, the report should display calculations
and analysis, charts, and figures, and at times significant raw data. Fourthly,
and most importantly, it should inform the reader of the particular results,
their significance, conclusions which may be drawn from them, and the
justification of such conclusions.
There are in general five major parts to a report even if a formal distinction
is not made between them. These are as follows:
1. Objective: The abstract tells the reader, when he/she first starts
reading, what he/she will find in the report if he/she reads it through.
In one or two sentences it should summarize all topics presented in the
report.