# Estimating sample size in experimental and non-experimental settings

Published on August 30, 2022   31 min
• Review

#### A selection of talks on Methods

Please wait while the transcript is being prepared...
0:00
Hello, my name is Benjamin Leiby. I am a Professor of Biostatistics at Thomas Jefferson University. Today I will be introducing the basics of sample size calculations in experimental and non-experimental settings.
0:14
One of the questions statisticians are often asked is how many people do I need to study to be able to have a statistically valid conclusion? My objectives today are to address this question and to review the basics of statistical hypothesis testing, to define Type 1 and Type 2 errors, to show what affects power and sample size, to demonstrate common power and sample size calculations. In this talk, I'm going to focus on two-group comparisons. For example, comparing people who receive a treatment to those who do not. I will be discussing this in the context of biomedical applications although the concepts can be applied to any setting where two groups are being compared.
0:55
First, let's review the basics of statistical hypothesis testing. A foundational distinction in statistical thinking is between the population and the sample. The population is the large, universal set of all people, or animals, or things, about which we wish to answer a question. For example, all people with lung cancer, or all people with stage three lung cancer. The population can be broad or specific, but it is not able to be completely measured or completely studied. In order to answer questions about the population, we study a sample from that population. We want to draw the sample well, so that we can actually make a conclusion about the population. Part of having a good sample is how it is drawn so that it is representative of the population. The other important part, which is the subject of today's talk, is the size of the sample. We want the sample to be big enough to have a precise estimate. We want to balance that against the costs of doing research. A study that is too large will cost more financially, but also could expose patients to risk unnecessarily. Based on the data we collect from the sample, we will test a hypothesis and then using the results of that test that we perform with the data from the sample, we draw our conclusions about the population. This process is called statistical testing or statistical inference.

Quiz available with full talk access. Request Free Trial or Login.

Hide