Part of teaching data analytics involves giving students an idea of the steps involved in performing analytical work. Here is my take on the steps. Please offer your comments below. The goal is to prepare young people to do better work when they go out into the field.
1. Know Your Client
On whose behalf am I performing this analysis?
The first step in a data analysis job is pinning down the identity of your client. Ultimately, data analysis is about creating information. If you are going to do that well, you need some sense of your audience’s technical background, their practical interests, their preferred mode of reasoning or communicating, and much else.
2. Start With a Clear Understanding of the Question
What information does my client want? Why do they want this information? How do they plan to use it, and is this something that I want to be a part of? Can I answer it using the resources or tools at my disposal, or do I have a clear idea of which resources or tools need to be required to get the job done?
You do not want to take on an analysis job that lacks clear informational goals. You may have to work with a client to pin down those informational goals. Doing so requires that you identify factual information that would substantially influence your client’s practical decision-making choices. In addition, it is best to avoid (or at least avoid taking lead on) jobs that you can’t handle. It is probably also a good idea to avoid jobs that are not genuinely interested or influenced by your findings, or begin with an answer already in mind.
Many clients are not highly familiar with data analysis, and may not be highly numerate. It is your professional responsibility to ensure that they understand what data analysis can and cannot do.
3. What Am I Supposed to be Examining?
What are my objects of analysis? Which qualities, behaviors, or outcomes am I assessing? Why?
These kinds of questions point students towards working hard to pin down the research design that guides their study. Researchers should have a clear sense of whose personal qualities or behaviors need to be examined, and exactly what about them is relevant to the client’s decision-making problem or goal. If done well, the researcher should have a clear sense of the theoretical propositions or concepts that are most crucial to the client’s goals, and how to find analyzable data to examine them.
4. Assess and Acquire Data
How can I get data on these units’ characteristics or behaviors? How do I get access to them? What is the quality of the data’s sample (i.e., how representative is the data)? Do these measures look reliable and valid?
All analyses should make some assessment of the data, using the tools that you acquired in your methods class. As the saying goes: Garbage In, Garbage Out. Don’t bother bringing a battery of high-powered analytical tools to a crap data set.
5. Clean and Prepare Data
Secure data, correct errors, refine measurements, assess missingness, and identify possible outliers.
This is the process of turning the raw, messy data that one often encounters into a clean, tidy set that is amenable to the analytical operations of your toolkit.
6. Implement Analytical Operations.
Implement analytical procedures designed to extract specific types of information from data.
Over the course of the semester, we will study a range of tools to extract information from data.
7. Interpret Analytical Operations
Convert statistical results into natural-language explanation, and assess implications to client’s thought processes.
The computer does the job of processing the data. Your job is to interface that output with the problem being confronted by your client. In the interest of making them a partner, it helps to develop the skill of translating complicated results into cognitively-accessible, but still rigorous and correct, explanations that everyone can understand.
You empower clients by giving them access to what you see in the statistical results. Win their confidence on this front, and they will work with you again in the future.
Finding ways to convey information in a way that resonates with client.
These are the skills that will keep you from being someone’s basement tech.
Photo Credit. United States Office Of Scientific Research And Development. National Defense Research Committee. Probability and Statistical Studies in Warfare Analysis. Washington, D.C.: Office of Scientific Research and Development, National Defense Research Committee, Applied Mathematics Panel, 1946. Pdf. https://www.loc.gov/item/2009655233/.