Filmina··9 min read

How to study statistics for social sciences: understand the data before the formula

Listen to article 10:55

If you're searching for how to study statistics for social sciences, here's the short answer: don't start with the formula, start with the question it answers. Every technique in the course (the mean, the confidence interval, chi-square, correlation) exists to answer something concrete about a group of people, and the exam almost always asks for exactly that: pick the tool, do the calculation and say what the result means in the language of your major.

What makes this course tricky in sociology, psychology, political science or communication is that it shows up for the people who least expected it. You picked a major built on theory and readings, and in your second year a course appears with slide decks full of symbols, a calculator and a program for analyzing data. Many treat it as an obstacle unrelated to the major, and it's the opposite: it's the course that teaches you to read a survey or a study without believing the first thing the headline says.

It isn't my major: I study computer engineering, and the statistics I took was a different flavor. What follows comes from looking closely at how the course is taught and graded in the social sciences, and from what's known about how people learn. It works for descriptive statistics, for inferential statistics and for research methods.

What the exam asks (and what it doesn't)

Before studying, it helps to know what's being graded, because it changes everything. In a statistics course for the social sciences there are usually three kinds of question:

Most people practice only the first, which is the most mechanical. But what lowers grades is almost always the second and the third. Knowing how to do the calculation without knowing what to say about it is worth half the answer.

The three questions that organize the whole course

A statistics lecture arrives with twenty concepts and it seems there's no way to organize them. There is one: in almost every topic the question is one of these three.

  1. How do I describe this data? Here you find frequency tables, charts, the mean, the median, the mode and the standard deviation. It's summarizing.
  2. How are two variables related? Here you find the crosstab, correlation, regression. It's comparing.
  3. Can I generalize what I saw in my sample to the whole population? Here you find the confidence interval, hypothesis tests, the p-value. It's inferring.

Make a sheet with those three questions and hang each technique under them as it comes up in class. When a new one arrives, the first question is "which of the three does it belong to?" That turns the syllabus from a list into a map.

For each technique, a four-line card

The card that works best for me for any concept with calculations is this:

The example from your major is what pays off most, because it gives the symbol a meaning. If you study psychology, the two-group t-test is "do the people who took the workshop score differently from those who didn't?" If you study communication, chi-square is "is the outlet people use to get their news related to their age?" The formula is the same, but now you know what it's for.

The costliest mistake: the p-value

Almost all the inferential statistics in the course ends in a p-value, and it's the thing most often misunderstood. It tells you how surprising your data would be if there were no real effect. It doesn't give the probability that your hypothesis is true, or tell you the effect is large or important. A small p-value with a huge sample can come from a tiny difference nobody cares about.

One way to use that on an exam: never write just "it's significant". Write a full sentence with the variable and the group: "the difference in study hours between students who work and those who don't is statistically significant, though we would need to look at its size". And remember the rule every professor repeats: correlation is not causation. In courses like sociology they ask for it almost every time in the interpretation part.

Practice problems blank, and mixed

In a course with calculations, reading the theory is the easy part and doing problems is what's graded. A very well established finding backs it up: testing yourself works better than rereading, and rereading is among the techniques that help least according to a large review. In statistics, testing yourself means solving the problem without looking at the solution, even if it goes wrong.

A routine that works for each topic:

  1. Cover the worked example from class and redo it from the prompt.
  2. Compare step by step and note which line you got wrong and why.
  3. Do one with different data, then one where something changes (another sample, a different variable).
  4. Go back to the ones you missed two days later, not the ones that already come out right.

Then mix them. Textbooks group the problems by topic, but on the exam they come shuffled and the hard part is recognizing which is which. In a classic study, students who practiced mixed types of problems did much better on the final test even though they did worse during practice. Make a sheet with one problem of each technique and solve it without knowing which chapter each comes from. I covered it in more detail in interleaved practice.

One full unit, with an example

Imagine that in the course you have to compare how many hours per week students who work and students who don't spend studying, with a survey of two hundred people. This is how I'd go step by step.

First, the question: is there a difference between two groups? It's a comparing question. Second, describe each group on its own: the mean hours and the standard deviation of each, and a box plot or histogram to see how they spread. Third, the technique: comparing the means of two independent groups calls for a t-test, as long as the data look enough like a normal distribution; if not, there are alternatives your instructor will give you. Fourth, the result: the statistic, the p-value and the confidence interval of the difference. Fifth, the sentence: what you conclude in words, what you can't conclude (that working causes less studying, for example) and what other variable might be mixed in.

In the social sciences, a number without its interpretation sentence is half an answer.

The order doesn't change with the technique: question, description, tool, result, sentence. With a sheet like that for each unit, you have the structure of an exam answer.

The data program is not the enemy

Many courses use a program (SPSS, R, Jamovi or even a spreadsheet) so you don't have to do the big calculations by hand. The trap is learning to click buttons without understanding what comes out. My advice: before running an analysis, write down what you expect to see and what each number in the output would mean. Then compare. If the program gives you a p-value, you should be able to say what it says and what it doesn't. And practice with a real or simulated dataset on your own topic, a survey, not only the textbook examples: it forces you to decide which variable is which.

The course slides, and some help

Statistics lectures have formulas, tables and charts, and what the professor says about each step isn't written down. For that part you can use Filmina, a web app that explains your slides out loud, one by one, pointing at the part it's talking about. It looks at each slide as an image too, so it explains the charts and tables together with the text, and if a formula doesn't add up you ask Mina, the chat that answers with your material (more in how it explains each slide). It has limits: it needs a connection and an account, the free plan has monthly minute caps (the plans are here), and a scanned PDF can't be read: if you have the original, upload that. The problems you have to solve yourself. For the general method with formulas, see how to understand slides with formulas.

A plan for the weeks before the exam

Spreading your reviews over different days is also backed up: spacing improves what you remember compared with doing it all at once. I lay it out with a calendar in spaced repetition. If you study politics, see also how to study political science, where statistics shows up in the methods part.

Frequently asked questions

How do I study statistics if I'm bad at math?

Statistics for the social sciences uses basic algebra, percentages and square roots; the hard part is choosing the technique and reading the result, not the arithmetic. Start with the question each technique answers, do problems blank and practice the interpretation sentence.

What is a p-value, in plain words?

It's how unusual your data would be if there were no real effect. A small p-value doesn't prove your hypothesis is true or that the effect is big: it only says the data are hard to explain by chance.

How many hours should I study statistics?

It depends on the course and your math background, but it pays off more to split them into several short sessions a week, with problems, than into one long one. Every session should include blank problems, not just reading.

What is statistics used for in the social sciences?

It's used to read surveys and studies with judgment and to know what you can conclude from a sample and what you can't. It's the base of research methods and of almost any work with data about people.

Keep reading

·6 min

How to study art history: artwork cards, styles and image analysis

One card per artwork, a timeline of styles, a three-step way to analyze an image and your readings in one table: how to study art history in college.

·8 min

How to study finance: from the time value of money to risk, with numbers in hand

Finance isn't studied by reading: time value of money, risk and valuation with worked problems, and every formula explained with its why.

·7 min

How to study public administration: rules, policies and cases

One sheet per concept with a real case, the policy cycle as a diagram and laws read with an example: how to study public administration in college.

Let's start
studying

Upload your next lecture and Mina does the rest: explains it, sums it up, quizzes you.