Anda di halaman 1dari 13

Error Analysis Tutorial

Welcome to the Error Analysis Tutorial. This tutorial will help you master the error analysis in the first-year, college physics laboratory. Error analysis may seem tedious; however, without proper error analysis, no valid scientific conclusions can be drawn. In fact, as the picture below illustrates, bad things can happen if error analysis is ignored. Since there is no way to avoid error analysis, it is best to learn how to do it right. After going through this tutorial not only will you know how to do it right, you might even find error analysis easy! The tutorial is organized in five chapters. Contents Basic Ideas How to Estimate Errors How to Report Errors Doing Calculations with Errors Random vs. Systematic Errors

Chapter 1 introduces error in the scientific sense of the word and motivates error analysis. Chapter 2 explains how to estimate errors when taking measurements. Chapter 3 discusses significant digits and relative error. Chapter 4 deals with error propagation in calculations. Chapter 5 explains the difference between two types of error.

1. Basic Ideas
This chapter introduces error analysis at the intuitive level. It is organized in three sections: 1.1. Errors as Uncertainties 1.2. Why Should We Care About Errors? 1.3. Dominant Error 1.1. Errors as Uncertainties

Before we go into details of error analysis it is important to understand the meaning of error in science. Error in a scientific measurement usually does not mean a mistake or blunder. Instead, the terms "error" and "uncertainty" both refer to unavoidable imprecision in measurements.

Of course, not all measurements have errors. If asked how many people there are in a room, one can usually give an exact number as an answer. However, if we want to know how many atoms there are in a room, giving an exact answer is nearly impossible, as the animation below illustrates.

In your laboratory you will deal mostly with measurements of the second kind. Since you will not be able to measure things with arbitrarily high precision, you should know how to quantify the imprecision of your results. 1.2. Why Should We Care About Errors?

You might wonder why physicists pay so much attention to errors. The reason is simple: the failure to specify the error for a given measurement can have serious consequences in science and in real life.

Consider the following situation. You have finally found an apartment in Manhattan which you can afford.

Filled with joy, you want to order a bed for your new bedroom. A friend of yours, who has no scientific background, tells you that he has measured the width of your bedroom and found it to be 2 meters. So you order a king size bed which is 198 cm wide. When it is delivered, it turns out that it doesn't fit into your bedroom. Now you measure your bedroom yourself and find that it is in fact 197

cm wide. No wonder the bed doesn't fit. So how could your friend's measurement have been so wrong? Well, it wasn't necessarily. Rather, he failed to tell you that since he used a short ruler, he could only measure it to a precision of plus or minus 5 cm. If you had the information about the large error in his measurement, you probably would have wanted to double-check his result before placing the bed order.

Needless to say, similar sloppiness in the engineering or medicine can have fatal consequences.

1.3. Dominant Error

In your lab more often than not, you will combine measurements of different quantities to obtain the final quantity you want. Let's assume we know the error associated with each individual quantity. The question is how to find the error of the combined quantity.

We will deal with this question in detail in a later section. However, it is very helpful to understand the concept of dominant error before doing any calculations. The idea of dominant error is very simple. Consider the two supermassive black holes first spotted in the NGC 6240 galaxy by NASA's Chandra X-ray observatory on July 29, 2001.

Two black holes in the NGC 6240 galaxy. Image credit: NASA/CXC/MPE

How far from Earth (in the Milky Way galaxy) is this pair of black holes (in the NGC 6240 galaxy)? The usual way of calculating large cosmic distances is to apply Hubble's Law, which states that the velocity with which two galaxies move away from each other is proportional to the distance between the galaxies. Algebraically we express Hubble's Law as

V=HD

Thus, if we know the velocity V of NGC 6240 relative to the earth and the so-called Hubble constant H, we can calculate the distance D between the Earth and the pair of black holes. On Earth, we can measure the velocity V quite precisely. However, we know the Hubble constant H only with a precision of about 10%. Therefore, the error estimate on the distance D is almost entirely

determined by the large error of the Hubble constant, and we can safely ignore the small error of the velocity V.

2. How to Estimate Errors

How does one actually give a numerical value for the error in a measurement? The answer to this question is in this chapter. As you will see, giving an error estimate for simple measurements is easy. The chapter consists of five sections: 2.1. Errors when Reading Scales 2.2. Errors of Digital Instruments 2.3. Standard Deviation 2.4. Histograms 2.5. Exercises

2.1. Errors when Reading Scales

The most common measurements in the lab are done with devices that have a marked scale. Let's look at an example. In one of your first experiments you will measure the length of a pendulum from the pivot point to the center of the mass attached to its end point. Let's first simplify the situation a little bit. We will measure the length of the pendulum from the pivot point to the visible end of the mass. The situation is schematically shown in the figure below

(numbers are in centimeters)

Even using this idealized, zoomed-in picture, we cannot tell for sure whether the length to the end of the mass is 128.89 cm or 128.88 cm. However, it is certainly closer to 128.9 cm than to 128.8 cm or 129.0 cm. Thus we can state with absolute confidence that the length L is

We call the first term, 128.9 cm, the "central value" and the second term, 0.1 cm, the "error" or "uncertainty".

If pressed, we could get a little bit better precision from the picture. However, in a real situation, the precision of 0.1 cm for measurements done with the centimeter ruler is as good as you can get.

2.2. Errors of Digital Instruments

Many modern laboratory instruments use digital displays. For example, you can measure weights using an electronic scale. It has the advantage that you don't have to judge what is the central value of your measurement. It will give you the central value. All you have to do is write it down.

Now, what is the error of this seemingly "precise" measurement? All instruments, no matter how sophisticated, have a limit to their precision. Usually, the manufacturer of the instrument will specify the error. (Most of the time it is written on the back of the instrument.) However, if neither you nor your laboratory instructor can find this information, use the following rule of thumb: The error of an electronic device is usually half of the last precision digit. The following example should make it clear. Suppose you measured your weight with the electronic device pictured above. The result is 138.2 lbs. What is the error? If you can't find the manufacturer information about the error, you notice that the resolution of your measurement is 0.1 lb. In other words, the device claims that it can give you a more precise value than 138 lbs, but it doesn't tell you whether your weight is 138.15 lbs or 138.25 lbs. Thus we can assume that the error of our measurement is 0.05 lb. And our answer is 138.2 lbs 0.05 lb.

2.3. Standard Deviation

Not all measurements are done with instruments whose error can be reliably estimated. A classic example is the measuring of time intervals using a stopwatch. Of course, there will be a read-off error as discussed in the previous sections. However, that error will be negligible compared to the dominant error, the one coming from the fact that we, human beings, serve as the main measuring device in this case. Our individual reaction time in starting and stopping the watch will be by far the

major source of imprecision. Since humans don't have built-in digital displays or markings, how do we estimate this dominant error?

The solution to this problem is to repeat the measurement many times. Then the average of our results is likely to be closer to the true value than a single measurement would be. For instance, suppose you measure the oscillation period of a pendulum with a stopwatch five times.

You obtain the following table:

Our best estimate for the oscillation period is the average of the five measured values:

Note that N in the general formula stands for the number of values you average.

Now, what is the error of our measurement? One possibility is to take the difference between the most extreme value and the average. In our case the maximum deviation is ( 3.9 s - 3.6 s ) = 0.3 s. If we quote 0.3 s as an error we can be very confident that if we repeat the measurement again we will find a value within this error of our average result.

The trouble with this method is that it overestimates the error. After all, we are not interested in the maximum deviation from our best estimate. We are much more interested in the average deviation from our best estimate. So should we just average the differences from our measured values to our best estimate? Let's try:

Clearly, the average of deviations cannot be used as the error estimate, since it gives us zero. In fact, the definition of the average ensures that the average deviation is always zero for any set of measurements. It is so because the deviations with positive sign are always canceled by the deviations with negative sign. Can't we get rid of the negative signs? We can. If we square our deviations, all numbers will be positive, so we'll never get zero1. We should then not forget to take the square root since our error should have the same units as our measured value. Thus we arrive at the famous standard deviation formula2

The standard deviation tells us exactly what we were looking for. It tells us what the average spread of experimental results is about the mean value. Now we can write our final answer for the oscillation period of the pendulum:

What if we can't repeat the measurement? The error estimation in that case becomes a difficult subject, one we won't go into in this tutorial. In your laboratory, the majority of relevant measurements are easily repeatable.

2.4. Histograms

In the previous section we've introduced the idea of obtaining a reliable error estimate by repeating measurements many times. This idea is so important in science and technology that an entire branch of mathematics, statistical analysis, is devoted to it. It is not our goal here to make you experts in statistical analysis, but it's worthwhile to learn a little bit about it.

Suppose we continue to take measurements of the oscillation period of the pendulum, as discussed in the previous section, until we have 10 measurements (in seconds):

3.9, 3.5, 3.7, 3.4, 3.5, 3.6, 3.7, 3.6, 3.8, 3.6

Let's count how many times each number occurs.

We can also plot our results as a bar histogram.

We can also plot our results as a bar histogram.

Note that the shape of the bar histogram resembles a bell curve.

If we had taken not ten but many more measurements, the bar histogram would resemble the bell curve even more1. In fact, for very large numbers of measurements the bar histogram would be indistinguishable from a continuous curve. This curve is called the limiting distribution. For most of the experiments in the lab the limiting distribution will be a bell curve, which is also known as a normal or Gaussian distribution.

The nice thing about histograms is that we can see the mean value of our measurements. In our example here, it's clearly very near 3.6 s. Can we find the standard deviation from a histogram?

First, let's calculate the standard deviation using the formula in the previous section. We obtain 0.14 s. Let's indicate this in our histogram.

We see that the points at 3.5, 3.6, 3.7 are within the range defined by the standard deviation. Note that in 7 out of 10 measurements, we obtained one of those three points. In other words, they account for 70% of our measurements. This is not a coincidence. It is a property of the normal distribution that 68% of the measurements lie within one standard deviation on either side of the mean. Since our distribution comes close to a normal distribution, our result 70% is close to the theoretical value of 68%.

By reversing the argument, we can find the standard deviation from the histogram. Given a histogram, we can find a region centered at the mean value, which includes 68% (about 2/3) of the measurements. Then, because of the property stated above, the width of this region will be 2 standard deviations. (One standard deviation for each side from the mean.) So in our example, suppose we don't know the formula for standard deviation. By looking at the histogram, we see that 7 out of 10 measurements, which is a good approximation to 2/3, are in the region between 3.45 s and 3.75 s. The width of the region is 0.3 s. Therefore the standard deviation is 0.15 s. Compare this to 0.14 s obtained via the formula. They are very close.

In the lab, in order to find the standard deviation, you can use either the algebraic formula discussed in the previous section or the 2/3 rule discussed here. The difference between the two methods can be considered negligible for the purposes of the lab.

2.5. Exercises

Usus magister est optimus. In other words, experience is the best teacher. Below are exercises covering the material in this chapter. After you make your selections, click "Submit" to see the answers. In a ballistic pendulum experiment, suppose the digital timer shows 0.02 s for the time of flight of the projectile. The manufacturer information about the precision of the timer is nowhere to be found. What error would you quote on your measurement? 0.002 s 0.005 s 0.01 s

In a projectile motion experiment suppose you have the following series of measurements of the distance x traveled by the projectile: 30 cm, 32 cm, 29 cm, 28 cm, 31 cm. The mean value is therefore 30 cm. What is the error (standard deviation) of the result? 0.8 cm 1 cm 1.4 cm 2 cm

Suppose you counted the number of people in line each time you came to your favorite coffee shop last month. Here are the results of your investigation.

What is the error determined by the 2/3 rule? 1.7 2 2.5 3 3.5

Anda mungkin juga menyukai