Accuracy vs Precision
Spot on! You hit the nail on the head! Bullseye! You've heard these statements if you correctly guessed the right answer to a question. Just like in that quiz, accuracy of scientific data refers to how close a measurement is to the 'true' value.
Let's take that bullseye for example. If you had a target that you were throwing darts at, and you hit the very center of the target each time, you would be very accurate because that is the 'true' value you are trying to achieve. But if you hit far outside the center of the target each time, you would be inaccurate because you are not near the 'true' value. The farther you are from the center, the less accurate your throws.
The difference between accuracy vs precision can be illustrated with a game of darts.
 |
Accuracy is very important in scientific data collection. For example, you may wish to measure the volume of a certain chemical in your experiment. If the actual volume was 60 ml but you measured 75 ml, you would not have a very accurate value because it is not close to the 'true' value of 60 ml. But if you measured 59 ml you might consider this an accurate value since it is so close to the 'true' volume.
Accuracy is important when collecting scientific data.
 |
Let's go back to your target for a minute. This time, when you throw the darts, you are nowhere near the center, but at least you hit the same spot every time. In this case, we can say that while you were not accurate, you were precise. This is because precision, which is often called reproducibility or repeatability, is the agreement of repeated measures.
So, while accuracy refers to how close you are to your true value, precision refers to how often you get the same measurement under the same conditions. If, for example, you were to measure the volume of your chemical ten times, and each time you measured 75 ml (when the true value is still 60 ml), you would still not be very accurate. The good news, though, is that you were precise because you got the same value each time you took a measurement.
As with the examples we just saw, you can be accurate without being precise and precise without being accurate. But you can also be both accurate and precise (getting close to the 'true' value each time) or neither accurate nor precise (being far from the 'true' measurement and at a different value each time).
Types of Error
Ideally, you want to be both accurate and precise. But because neither humans nor the instruments we use to take measurements are perfect, any measurement is prone to error in some form. Human error includes any basic human mistake: spilling a substance, dropping equipment, forgetting to turn off the drying oven, etc. These errors can and will happen, no matter how avoidable they may seem in hindsight!
Any errors that are not directly attributed to human 'oopses' fall under two categories. The first is random error, which is error that occurs randomly in space and time. This type of error occurs periodically and in no specific pattern and, therefore, gives you imprecise results. Random errors are unavoidable and are just part of every dataset, but they are just as likely to occur in a positive or negative direction. This means that theoretically, after you take enough measurements, the random errors would essentially cancel themselves out, getting you closer to your 'true' value. Repetition works in your favor here!
The second type of non-human error is systematic error, which is error associated with the instruments. Perhaps your pipette is not dispensing the correct amount of fluid, or your scales are not calibrated correctly. In this case, you would have the same error each time because your instrumentation is off by the same amount for each measurement. In the end, all your measurements would be too high or too low, leading to precise, but inaccurate, data.
One kind of systematic error you may encounter is drift. This is when an instrument gradually changes over time. Drift often occurs in instruments that record continuously, like detectors. Each type of instrument will have a different amount of drift - for example, some bat detectors drift about eight minutes per year, while some fish detectors drift a few seconds each month. A few minutes or seconds may not sound like a lot, but if you're trying to use those detection data to calculate swimming speeds in meters per second, you need the seconds to be the right ones! As you can see, this is a really important type of error to account for. Most likely, though, the instrument manufacturer will tell you how much drift will occur and how often to re-calibrate your machines in order to avoid it.
Lesson Summary
Reporting data is an important step in scientific investigations. Your data need to be credible, though, because others may base new experiments off your results, policy decisions may be influenced and in general, you want to be seen as a reputable scientist.
Examining your data for accuracy and precision helps give credibility to your results and experiment. Accuracy refers to how close a measurement is to the 'true' value, and precision, often called reproducibility or repeatability, is the agreement of repeated measurements. Hitting the center of the target means you are accurate. But hitting the center of the target over and over again is even better because now you are both accurate and precise.
Error will occur in any data set, especially random error. These errors that occur randomly in space and time are both unavoidable and unpredictable. However, with enough replication, they will essentially balance themselves out. Systematic errors on the other hand, are associated with instruments and can be prevented with preventative steps like proper calibration.
Learning Outcomes
Determine your ability to do the following when you've reviewed the video:
- Understand the importance of reporting credible data
- Indicate the roles that accuracy and precision play in the process of producing credible data
- Specify the types of errors that can occur in any data set
