Statistics in Research

Regardless of the experimental design and question, research involves measuring and interpreting variables. In an experiment, we manipulate one or more variables, measure at least one other, and control all the rest. Research design is concerned primarily with establishing rules for manipulation and control of the variables, while the discipline of statistics is concerned with measurement and interpretation of variables. More specifically statistics serves three purposes in research: description, inference, and communication.

Descriptive statistics make the results of research projects understandable. It's usually very difficult to draw conclusions from a mass of raw numbers. Statistics offer ways to condense large volumes of raw data so that their important features stand out clearly. Without descriptive statistics we would have to rely on hunches and impressions derived from an inefficient look at the data.

Inference. We're rarely interested in only the particular set of data that comes from an experiment. Rather, we hope that the results of one project will apply to other similar situations and groups of subjects. The particular group of observations in a research project is a sample of the possible samples that we didn't observe. All the possible samples would make up a population and it is this population that is normally of interest in research. Since it is usually impossible to measure the entire population, using samples from that population allows us to make statements about the population. Inferential methods allow us to make estimates of parameters (characteristics of populations) on the basis of sample statistics (characteristics of sample). Inference fills an important role in science - it gives us a way to have confidence in the conclusions that we draw from a sample. Since science tries to understand the laws that govern the external world and since we want those laws to be as broadly applicable as possible, we don't want the results of a single experiment to be limited to only the participants in a single project. The methods of inferential statistics allow us with a means of generalizing form the sample to the entire population.