One of the most important things to remember about science is that the scientific method is empirical. Empirical means based on observation. In addition to observing, scientists make systematic observations. By systematic, I mean that the scientist observes with a plan, and the plan is designed to ensure objectivity. Objectivity means that the scientist takes steps to record facts that are not colored by emotion and personal prejudice. Recording observations as numerical measurements greatly aid the researcher in maintaining objectivity.
Empirical means based on observation rather than other methods of knowing, such as reliance on authority or “common sense.”
This method presents a problem. All of these observations result in huge amounts of numbers. We can organize those numbers into a table form such as a spreadsheet, but there will still be pages and pages of them. Information (data) presented in this way is meaningless! The human mind simply cannot wrap itself around large amounts of numbers like that and draw anything meaningful from them. We need to organize and simplify the data. Organizing, simplifying, and summarizing data is a primary function of descriptive statistics.
Descriptive statistics are a family of statistical methods that organize, simplify, and summarize data.
Within the scope of empirical research, there are several strategies that the social scientist can use to answer questions. An important distinction is between whether the research is descriptive, relational, or experimental.
Descriptive research seeks to describe the characteristics of a particular social phenomenon. Researchers doing descriptive research ask, “What is the state of things?” Public opinion polls are an example of descriptive research. They answer questions about how people feel about particular issues. To be merely descriptive, the researcher treats these opinions independently and does not consider how these relate to other variables and does not offer up any potential causes.
Relational research seeks to understand how two or more variables are related. If our pollster in the above example examined how religious preference was related to whether a voter was going to vote Democrat or Republican, the study would be relational. This is because the researcher is examining the relationship between religious preference and voting preference.
Experimental research seeks to make causal statements about how the social world works. That is, the researcher wants to make cause and effect statements. Did a university’s alcohol awareness program (cause) reduce incidents of alcohol-related crime on campus (effect)? Note that the confidence with which a researcher can make such statements relies heavily on good experimental design, which is beyond the scope of this text. We will focus strictly on the statistical concerns of good research design.
Note that these types of research are not mutually exclusive. In the practical world of research, they tend to work in concert, building on each other. A researcher conducting an experiment to demonstrate that one variable causes another will no doubt want to describe both variables and explain how they are related as well. In other words, these divisions are somewhat arbitrary, but they are useful because each has a set of statistical techniques that goes along with it. Descriptive statistics are heavily used in descriptive research, and inferential statistics are used heavily in experimental research. Relational research uses a family of correlational statistics that can be both descriptive and inferential depending on how they are treated.
Descriptive statics, then, seek to summarize and explain the characteristics of a variable. The other major branch of statistics, inferential statistics, lets the researcher make generalizations about a population given information from a sample. Generalizations are general statements obtained by inference from specific cases.
Generalizations are general concepts that come from inferences from specific cases.
An inference is a conclusion that is based on facts and reasoning. We can gather from these definitions that the purpose of inferential statistics is to let us make general statements about populations based on information that we have gathered from samples.
Inferential statistics is the branch of statistics that uses sample data to reach conclusions (make inferences about) populations.
To understand the jargon of inferential statistics, it is helpful to understand how social scientists answer most questions. It is important to understand that scientists are generally interested in populations. A population is the entire group of people that a researcher is interested in making statements about. A population can be very small, such as female Supreme Court justices, or it can be very large, such as every registered voter in the United States.
A population is the entire group of interest in a research study.
A number that describes the characteristics of a population is called a parameter.
When a researcher collects information about an entire population, it is called a census. Numbers that describe characteristics of populations are known as parameters. For example, when the Bureau of the Census reports the median household income of Americans, that is a parameter because it is a number that came from data collected on the entire population.
A census is information collected from the entire population of interest.
When populations are large, it becomes unfeasible to collect data for every person. There simply are not enough resources (human resources, time, money, etc.) to conduct a census. When this is the case, researchers use samples. A sample is a subset of a population that is used to answer questions about the population from which it was drawn. Numbers that describe characteristics of samples are called statistics.
A sample is a subset of a population that is used to answer questions about the population.
A number that describes the characteristics of a sample is called a statistic.
Social scientists using inferential statistics start with a hypothesis and look to see if data gathered from systematic observations are consistent with the hypothesis. Many of these techniques are complex and are best left to computers. For now, we will define a hypothesis as an educated guess as to how some social phenomenon occurs. (We will delve more deeply into hypotheses when we get to the sections on inferential statistics).
A hypothesis is an educated guess as to how some social phenomenon occurs.
Even if you do not plan to conduct research, you must be an intelligent consumer of research to assume a leadership role in today’s data-rich world. The social science professions, even the applied ones such as social work and criminal justice, depend on social scientific research to determine best practices, prepare grant applications, conduct program evaluations, and many other management tasks. In every arena of public management, there is a growing demand for accountability. Accountability in a narrow but important sense means that tax dollars are spent on programs that work and work well. The public demands objective evaluations of programs and policies. Such evaluations are performed using the same techniques that social scientists use to answer other questions.
Never forget, then, that the ultimate goal of statistics is to explain real-world events. This is a two-part process that calls for the objective application of the statistical methods that we’ll learn in this little book, and it requires subjective judgments as to what is good, what is right, and what is meaningful. Ultimately, science in general and statistical methods, in general, are only useful in informing the subjective judgments of decision makers. Science is not a tool for making ethical decisions, but it is a tool for informing ethical decisions makers. Statistics is a scientific tool, but the interpretation and application of that objective information are often more of an art than a science. That is why a major focus of this book is understanding why we compute a particular statistic, and what exactly the results mean.
Key Terms
Statistics, Descriptive Statistic, Average, Empirical, Objectivity, Data, Systematic Observation, Inferential Statistic, Sample, Population, Percentage, Margin of Error, Census, Parameter, Generalization
This work is licensed under an Open Educational Resource-Quality Master Source (OER-QMS) License.
Last Modified: 06/28/2018