Chi-Square in JASP: Step-by-Step Tutorial with a Psychology Example (2026)

If you need to run a chi-square test and don't want to wrestle with code or pay for an SPSS license, JASP is probably your best bet. It's free, open-source, and has a visual interface that shows you results in real time as you set up the analysis. Plus, it gives you something neither SPSS nor R offer out of the box: the Bayes Factor.

In this tutorial I walk you through the entire process: from opening your data in JASP to interpreting every output table and reporting results in APA 7 format. We'll use a clinical psychology example so you can see exactly what to expect.

When to use the chi-square test of independence

The chi-square test of independence (also called Pearson's chi-square) is used to assess whether two categorical variables are associated. In other words, whether the distribution of one variable changes depending on the categories of the other.

Typical scenarios in psychology:

• Assessing whether therapy type (CBT vs. psychodynamic) is associated with treatment outcome (improved vs. not improved).
• Determining whether gender is associated with the presence of a clinical diagnosis.
• Testing whether education level relates to therapeutic format preferences.

What chi-square does not do: it does not establish causality, it does not work with continuous variables, and it is not appropriate when expected frequencies are very low (for that, use Fisher's exact test -- and JASP shows it to you automatically when you need it).

Assumptions of the chi-square test

Before jumping into the analysis, check that your data meet these requirements:

The independence assumption is the most important one and cannot be tested statistically -- it depends on your study design. If each participant appears only once in the table, you're fine. If you have repeated measures, you need McNemar's test, not chi-square.

Worked example: our dataset

Let's use a concrete clinical example. A research team wants to assess whether therapy type (CBT vs. psychodynamic) is associated with treatment outcome (clinically significant improvement vs. no improvement) in a sample of N = 120 adult patients with major depressive disorder.

Your data can be in a .csv file with two columns (one per variable and one row per participant) or in an already aggregated contingency table format. JASP handles both, though working with participant-level data is more common.

Chi-square in JASP: step by step

Let's get to it. Here's the full analysis in JASP, from opening your data to having results ready for your paper.

Step 1: Open or import your data

Open JASP and load your data file. JASP accepts .csv, .sav (SPSS), .ods and other formats. Once loaded, you'll see your data in the left panel like a spreadsheet.

Before moving on, check that JASP classified your variables correctly. Click on each column name and make sure it shows as Nominal (not Scale). If JASP read them as numeric, change it right there. This matters because JASP will only let you use nominal or ordinal variables in the contingency tables module.

Step 2: Go to Frequencies > Contingency Tables

In the top toolbar, click the Frequencies icon (it looks like a frequency table). A dropdown opens. Select Contingency Tables.

This opens the configuration panel on the left and a results panel (empty for now) on the right. As you configure the analysis, results will appear in real time. No need to click a "Run" button.

Step 3: Assign variables to rows and columns

In the configuration panel you'll see your variables listed on the left. Drag one variable to the Rows box and the other to the Columns box.

In our example: Therapy type to Rows and Treatment outcome to Columns (or the other way around -- the order doesn't affect the statistical result, only the table layout).

As soon as you drag the variables in, JASP shows you the basic contingency table in the right panel. Fast.

Step 4: Configure Statistics options

Below the variable area, in the Statistics section, check the options you need:

• X² (Chi-squared): enabled by default. This is the main test.
• Phi and Cramer's V: check this. It's your effect size measure, required by APA 7.
• Odds ratio (2x2 tables only): useful in clinical contexts. Gives you the odds ratio with its confidence interval.

You'll also see a Log odds ratio option. For most psychology analyses I work with, it's not needed, but if you're doing meta-analysis, it can be helpful. If you just need the odds ratio from a 2x2 table without opening JASP, our odds ratio calculator gives it to you directly, with the confidence interval and the relative risk.

Step 5: Configure Cells options

In the Cells section, enable these:

• Expected counts: essential. Shows expected frequencies in each cell, which you need to verify the assumption (all >= 5).
• Row percentages or Column percentages: make interpretation easier. They let you say things like "70% of CBT patients showed improvement" instead of working only with raw frequencies.

Results update instantly in the right panel. No "calculate" button to press.

Step 6: Reading the JASP output

JASP generates several tables in the results panel. Let's go through each one.

Contingency table

The first table shows observed and (if you enabled them) expected frequencies in each cell. This is where you verify the expected frequencies assumption.

In our example, all expected frequencies are 24.0 or 36.0. Well above 5. The assumption is met without issues.

Chi-Squared Tests table

This is the main table. Look for the X² row (Pearson's chi-square). It reports three things:

• Value: the chi-square statistic. In our case, 4.44.
• df: degrees of freedom. For a 2x2 table, always 1.
• p: the p-value. Here, .035.

Since p = .035 < .05, we reject the null hypothesis: the two variables are not independent. There is a statistically significant association between therapy type and treatment outcome.

You'll also see the N row, which simply confirms your total sample size (120).

Effect size table (Nominal)

If you checked Phi and Cramer's V, JASP shows a separate table with effect size measures:

• Phi: .192 (for 2x2 tables this equals Cramer's V).
• Cramer's V: .192.

According to Cohen's (1988) benchmarks, .19 falls between a small effect (.10) and a medium effect (.30) for df* = 1. The association is significant, but modest. This is useful information: it tells the reader that yes, there is an association, but it's not an overwhelming effect.

Step 7: The Bayesian option (Bayes Factor)

This is where JASP really sets itself apart. In the Statistics section, check the Bayesian contingency table test box. JASP will calculate the Bayes Factor (BF₁₀).

The BF₁₀ tells you how many times more likely the data are under the alternative hypothesis (there is an association) compared to the null (no association). For example:

This is especially valuable when your p-value hovers around .05 (like our .035). The Bayes Factor lets you quantify the evidence more precisely than a simple "significant / not significant" call. More and more psychology journals appreciate seeing it included.

Step 8: When JASP automatically shows Fisher's exact test

If any cell in your contingency table has an expected frequency below 5, JASP automatically includes Fisher's exact test in the output. You don't have to do anything special -- it appears as an additional row in the Chi-Squared Tests table.

In our example it doesn't apply (all expected frequencies are above 5), but it's good to know. If Fisher's test appears, report that value instead of Pearson's chi-square.

Interpreting the results step by step

Now that you have the full JASP output, let's interpret each piece in order. If yours shows something that is not covered here, paste the JASP output into this tool and it explains what each number means.

1. The contingency table

Compare observed frequencies against expected ones. Expected frequencies represent what you'd see if the variables were completely independent. The more observed values deviate from expected ones, the larger the chi-square statistic.

In our example: we'd expect 36 improvements in each group (if therapy type were irrelevant), but we observed 42 in CBT and only 30 in psychodynamic. That discrepancy is what chi-square quantifies.

2. The chi-square statistic, degrees of freedom and p-value

The chi-square statistic summarizes the total discrepancy between observed and expected frequencies. It is calculated as:

Degrees of freedom: (rows - 1) x (columns - 1). For a 2x2 table: df = 1.

Our result: X²(1, N = 120) = 4.44, p = .035. We reject the null hypothesis of independence.

3. Effect size

A significant p-value doesn't tell you how large the association is. That's what Phi and Cramer's V are for.

• Phi: used exclusively for 2x2 tables. Equivalent to a correlation between two dichotomous variables. Ranges from 0 to 1.
• Cramer's V: works for tables of any size. For 2x2 tables, it equals Phi.

Our V = .19 sits between small and medium. There's a statistical association, but it's not a large effect. Pretty typical for clinical psychology research, by the way.

When to use Fisher's exact test

Fisher's exact test is the alternative to chi-square when the expected frequencies assumption fails:

• More than 20% of cells have expected frequencies below 5.
• Any cell has an expected frequency below 1.
• Total N is very small (under 20-30).

The technical difference: chi-square uses an asymptotic approximation to the chi-square distribution, while Fisher calculates the exact probability. That's why Fisher is more accurate with small samples.

The good news in JASP: you don't have to choose manually. If your expected frequencies are low, JASP shows Fisher's test automatically alongside chi-square. Just look at the corresponding row in the results table and report that value instead.

Effect size: Phi vs. Cramer's V

Reporting only the p-value is insufficient. APA 7 requires you to include an effect size measure. For chi-square your options are:

• Phi (2x2 tables): the square root of chi-square divided by N. Identical to Cramer's V for 2x2 tables. Interpreted like a correlation: 0 = independence, 1 = perfect association.
• Cramer's V (tables of any size): the generalization of Phi for larger tables. Always ranges from 0 to 1 regardless of table dimensions.
• Odds Ratio (2x2 tables): if your table is 2x2, you can also report the odds ratio as a complementary measure. JASP calculates it directly if you check the box. Particularly useful in clinical contexts.

APA 7 reporting

Here's how to report chi-square following APA 7 guidelines. You need the statistic, degrees of freedom, N, p-value and effect size:

Key elements:

• Use X with a superscript 2 for the statistic (italicize if your word processor allows it).
• Degrees of freedom and N in parentheses.
• Exact p-value, not just "p < .05".
• Include Cramer's V or Phi.
• Describe the proportions to give substantive meaning to the result.

If you also obtained the Bayes Factor, you can add it at the end. It's not required by APA 7, but more and more journals appreciate it:

Common mistakes with chi-square

After reviewing hundreds of theses and manuscripts in my statistical consulting practice, these are the mistakes I see most often:

Quick reference

If you're working on your thesis or a manuscript and need help deciding whether chi-square is the right test, checking assumptions, or interpreting results, my statistical consulting service can review your analysis and make sure everything is correct before submission. A timely review saves costly back-and-forth with reviewers later.