By Luke Wilmshurst
For many graduate students, and even some faculty members, statistical analysis can be the least enjoyable part of preparing a study for publication. Recently, increasing attention toward criticism of the methods used to analyze data has created an environment that can sometimes be hostile or downright intimidating. Now, perhaps more than ever, it is crucial to have a strong understanding of how to administer the right tests, in the right way.
Fortunately, a great resource is available on the SPSP website, with content including course lecture notes, ‘how to’ guides for multiple statistical packages (SPSS, R, Stata etc), articles covering specific topics (e.g. structural equation modeling), and even links to external sources for data. This article aims to supplement the SPSP statistics webpage, by suggesting some additional resources, which meet needs that correspond to the following three questions.
- 1. Which statistical test should I use?
Suppose you are working on a study and, although things seem promising, you are not sure how to get started with analyzing the data. Often, university courses will define key concepts and provide formulas for a range of methods, emphasizing theory in great depth, while neglecting to make students comfortable applying these tests during their own research. A useful resource to help bridge this theory-practice divide is some sort of decision tree or table, to offer a simple way to identify which statistical test is appropriate, for a given data set.
One fantastic starting point is this one-page decision tree that provides step-by-step guidance on which test to use, in the form of a graphical flowchart. Tools like this do a great job of putting theory and calculations aside, temporarily, to help readers focus on getting the first steps right.
A variation on the flowchart above can be found in the form of a web-based table, which includes several dozen links to guidelines for how to carry out each of the statistical tests, using four different statistical packages (SPSS, Stata, SAS and R). Similar information, in the format of printable reference sheets can be found here, which includes a quick one-line summary of what each test accomplishes, and here which matches each test with a specific question it is able to answer (see page 5).
- 2. Where can I find resources to get more hands-on practice analyzing data?
Another problem can occur when statistics courses are overly theory-based, which are well suited to reinforce technical knowledge, but can neglect to give enough attention to creating a bridge that gives students practice applying them in contexts relevant to their own research needs.
One fun interactive tool is this fantastic web-based P-hacker app. While most researchers may be familiar with the concept of p-hacking, being able to play around with data on this site is a great way to gain a stronger understanding of exactly how sensitive the relationship is between manipulating data and what results are generated.
Online textbooks, lecture notes and power point slides can be another easily accessible way to brush up on a range of topics, but while this kind of resource can be helpful as a reference guide, they typically have one key drawback: a lack of interactivity, and no opportunity to interact with instructors or classmates when questions arise. One way to solve this problem is by registering for an online class. A broad range of MOOC courses are available for free via sites such as Coursera and edx. While some courses are offered only a few times per year, others have archived content that can be started at any time. These courses often make use of multimedia materials (e.g. videos, downloadable slides), with practice assignments and message boards that allow users to interact with classmates and TAs, who can help answer questions.
- 3. I have a specific question. Where can I find information and/or detailed guidance?
Message boards and blogs represent two types of free online resources that often touch upon specific, and sometimes obscure, topics, with crowdsourced responses that can help resolve knowledge gaps. Reddit is one of the larger message boards, and has dedicated mini sites including statistics, ask statistics, and psychology. Alternatively, a number of blogs can be useful, either for simplifying concepts or providing opportunities to engage in technical discussions of special topics (for example, see here, here, or here). However, one word of caution: the quality of responses can vary from post to post, so a little bit of good judgment can go a long way here. Finally, online publications specialized on data analysis, such as five thirty eight go beyond psychology, but offer some interesting articles linking statistics with topics ranging from sports to analyzing social trends, to political forecasting.
Did this article provide any helpful suggestions? Please feel free to reach out via email if you have any suggestions of great resources we missed, or suggestions for topics to be covered in future issues.