Reading in Science
Observation, Measurement, and Data
The key scientific habits of mind—curiosity and skepticism—are focused on data, on obtaining it and understanding what it means. When scientists talk about evidence, they typically mean collections of data. To read science well, students need to understand some fundamentals about data, including its limits and shortcomings.
Science is concerned with observing and measuring natural phenomena—in some cases directly, and in other cases, resulting from the manipulations of the investigator in the field or in the lab. In practice, scientists will call nearly any measurement they make an experiment, even if they aren’t manipulating anything. As scientists read both text and graphics, they are constantly thinking about data and evidence.
Important aspects of scientific data and evidence include the following:
- Observations and qualitative descriptions. Qualitative <p><strong>qualitative</strong><br /> Data and evidence based on observations that do not include numerical measurements and are not, strictly speaking, subject to statistical analysis. Qualitative data are likely to be subjective, meaning different observers of the same phenomena might differ in what they record. Qualitative data can be made less subjective by repeated observations by multiple researchers. Single occurrence observations, referred to as anecdotal, are generally avoided.</p> observations are somewhat rare in the modern scientific literature, but do exist. For example, a research report might contain a description of animal behavior, or a chemist might record a color change and odor associated with a chemical reaction. There isn’t a way to do statistical analysis of observations without first making them quantitative. The most common source of error is misjudgment on the part of the observer. Repeated observations, careful notes, and technology to record and store observations are common ways to minimize errors.
- Counting categorical data. Some data collection involves counting without any measurement. The researcher must apply criteria for whether the observation fits in the category. For example, in counting the number of cars going past a school at different times of day to get a measure of traffic, the decision must be made on whether to include motorcycles. Many sophisticated research reports include counted data; for example, the number of particles emitted in a high-energy physics experiment or the number of mitochondria found in a cell. Statistical tests can be applied to counts, also called frequency data. Errors can be made in categorizing or in counting; the best way to mitigate this is to have a large sample size and more than one person counting.
- Quantitative <p><strong>quantitative</strong><br /> Data that are the result of counting and measuring and are thus subject to statistical tests of the shape, distribution, and significance. However, quantitative data are not intrinsically more dependable than qualitative data.</p> measurements. Beyond simply counting, careful measurements are a typical aspect of many experiments. Physical dimensions of size and weight—and also things like energy produced, or speed—may be measured, and it is important to keep track of the unit of measure. Two important aspects of measurement are precision and accuracy. If a metric ruler is being used to measure length, the precision <p><strong>precision</strong><br /> A term that describes the resolution of the recording instruments used in collecting quantitative data. It is not possible to measure at a finer grain than the instrument is capable of, and data should not be reported beyond that resolution.</p> will be defined by the smallest increments marked on the ruler. Accuracy is a reflection of how consistently and effectively the ruler is being used as a measuring tool. If the researcher is sloppy and inconsistent, that will lead to inaccuracy. It is possible to use a high-precision tool in a very inaccurate way, and also to be highly accurate with a low-precision measuring tool.
- Defining and controlling variables. The classic laboratory experiment is designed to record how one thing changes (the dependent variable) in response to experimentally changing a single other factor (the independent variable) while attempting to keep all other factors constant. Designing a good experiment involves being able to incrementally alter the independent variable while making accurate and precise measurements of the dependent variable. Working scientists don’t use the terms dependent and independent variable in their research or in their writing. They are awkward terms and easily confused with one another; in practice. what’s being measured is obvious. Scientists are more concerned with what they call controls (see below).
- Comparison, controls, and sampling. Very accurate and precise measurements can be produced, but may be of limited value in drawing conclusions if a proper set of comparison or reference data hasn’t been identified. Reference or control information can come from published sources or from control experiments done by the researcher. For example, if a new fertilizer were being tested on plants, a typical control experiment would be to not treat half of an experimental field with the fertilizer in order to have a comparison for the half that did get the fertilizer.
Apply: If possible, work with one of your colleagues in mathematics to choose two or three data examples that would help introduce your students to data basics, including statistics. Invite your colleague to visit your class as a guest speaker, and treat the person like an outside expert. Students might interview the guest on his or her education and training and introduce the guest before the presentation. The goal is not to have a mini-course on statistics, but rather for students to think seriously about data analysis and to begin to see the important connection between math and science. You might choose real examples: perhaps a normal distribution, a histogram with error bars, or a scatter plot that is nearly linear.