What does it mean to make a statistical inference? As opposed to just reporting descriptive statistics for the data you collected from a sample, statistical inference is a process where you use data from a sample to describe properties of the distribution of data in the population. When you test a hypothesis, calculate a confidence interval, or estimate an effect size, you are making statistical inferences.
The first thing you need to know is that there are different approaches to making statistical inferences. In this course, we will discuss Frequentist, Likelihood, and Bayesian approaches. All these different approaches have strengths and weaknesses, as you will
discover in Lecture 1.1. It’s also useful to realize that within Frequentist statistics
(which defines the probability of the outcome of a study in light of its
frequency in a very large number of repetitions of the study) there are two
different approaches to using p-values: One proposed by Ronald Fisher, and one
by Jerzy Neyman. People often use p-values in a hybrid approach that
randomly combines things from the Neyman-Pearson approach and the Fisherian approach, but this has been widely criticized.
This video is part of my free online course on statistics: www.coursera.o...
Can't get enough? Some suggestions for additional reading:
Lecture 1.1:
A great accessible overview of different perspectives to draw statistical inferences is provided by Zoltan Dienes (although it has Psychology in the title, it is really relevant for all empirical sciences using statistics):
Dienes, Z. (2008). Understanding psychology as a science: An introduction to scientific and statistical inference. Palgrave Macmillan.
A good overview of Frequentist statistics (with a strong focus on confidence intervals and estimation) that is full of educational information is provided in a book by Cumming:
Cumming, G. (2013). Understanding the new statistics: Effect sizes, confidence intervals, and meta-analysis. Routledge.
14 окт 2024
