The Chi-Squared Test: Rationale, Application and Critical Values - Worksheets, Questions and Revision
12 original exam-style questions - 3 pages of questions with a full mark scheme - free printable PDF.
The Chi-Squared Test: Rationale, Application and Critical Values revision guideWrite-in editionMore A Level Psychology practiceMock exams
This topic is chapter 12 of A Level Psychology Topic Workbook 4 and chapter 5 of A Level Psychology 10-Minute Tests 3.
7.2 The Chi-Squared Test: Rationale, Application and Critical Values
Question 1
- A test for association or difference between two categorical/nominal variables using frequency counts, typically applied to an independent-groups contingency table oeB1
- Answer: A test for association or difference between two categorical (nominal) variables using frequency counts, usually applied to data in an independent-groups contingency table.
Question 2
- critical value = 3.841 caoB1
- Answer: 3.841
Question 3
- df = (2 - 1) x (2 - 1) = 1 caoB1
- Answer: df = 1
Question 4
- explains that Chi-Squared uses counts of categories and would require arbitrary binning of continuous data which loses information oeM1
- names appropriate tests such as t-tests or ANOVA for comparing group means of continuous data (parametric tests) oeA1
- Answer: Chi-Squared uses categorical counts and would force arbitrary grouping of continuous data, losing information; for continuous interval/ratio data use t-tests or ANOVA to compare group means.
Question 5
- correct expected frequency 40.5 with working E = (row total 60 x column total 81) / 120 = 4860/120 = 40.5 caoB1
- Answer: Expected frequency = (60 x 81) / 120 = 4860 / 120 = 40.5.
Question 6
- formula given correctly as X2 = sum((O - E)2 / E) caoM1
- explains O = observed frequency in a cell and E = expected frequency in that cell under the null hypothesis oeA1
- Answer: X2 = sum((O - E)2 / E). O is the observed frequency in a cell and E is the expected frequency for that cell under the null hypothesis.
Question 7
- states df = (number of rows - 1) x (number of columns - 1) caoM1
- gives example in words: for a 2 by 3 table df = (2 - 1) x (3 - 1) = 1 x 2 = 2 oeA1
- Answer: df = (rows - 1) x (columns - 1). For a 2 by 3 table df = (2 - 1) x (3 - 1) = 2.
Question 8
- feature indicating Chi-Squared: both variables are nominal/categorical and recorded as counts/frequencies oeM1
- feature indicating Chi-Squared: independent groups design where participants are in separate categories (contingency table) oeM1
- feature favouring a t-test: the dependent variable is continuous (interval/ratio) and approximately normally distributed so group means are compared oeA1
- Answer: Chi-Squared: two nominal categorical variables recorded as frequencies, and an independent-groups design with a contingency table. A t-test is appropriate if the DV is continuous and approximately normally distributed so that group means are compared.
Question 9
- explains that one-tailed critical values test for an effect in a specified direction, while two-tailed test for any difference in either direction oeM1
- states that Chi-Squared tests of association use the upper-tail critical value (conventionally the two-tailed table is not relevant because Chi-Squared is always non-negative, so the standard critical values give the threshold for significance in the upper tail) oeA1
- Answer: One-tailed looks for an effect in a specified direction while two-tailed looks for any effect; for Chi-Squared association tests we use the upper-tail critical value (effectively the standard table threshold for the upper tail) since X2 is always non-negative.
Question 10
- limitation: Chi-Squared may be inaccurate when expected frequencies in one or more cells are too small (commonly when any expected < 5 for a 2x2 table) oeM1
- alternative: Fisher's exact test can be used for small samples/low expected frequencies oeA1
- Answer: Chi-Squared can be inaccurate when expected frequencies are small (often any expected < 5), so Fisher's exact test is an appropriate alternative for small samples or 2x2 tables.
Question 11
- Level 4 Knowledge of the Chi-Squared test and its use is accurate and well developed. Evaluation is thorough and balanced, considering methodological, statistical and practical issues, and is applied to psychological examples. There is effective, integrated use of strengths and limitations, and conclusions are clear and justified. Specialist terminology is used correctly.13-16
- Level 3 Knowledge is sound with some development. Evaluation addresses several relevant points, but may be less comprehensive or less well integrated. Application to psychological examples is present. Explanation is mostly clear and terminology is generally used correctly.9-12
- Level 2 Knowledge of the test is partial and evaluation is limited or superficial. Some relevant points are made but development and application to psychology is limited. The answer may lack clarity or organisation in places.5-8
- Level 1 Very limited knowledge and evaluation. Points may be simple or list-like and application to psychology is minimal or missing. Answer lacks structure and specialist terms are used incorrectly or not at all.1-4
- Level 0 No creditable content.0
- Indicative content:
- AO1: Description of what Chi-Squared tests: a non-parametric test for association or difference between categorical variables using observed versus expected frequencies in contingency tables.
- AO1: Formula and degrees of freedom: X^2 = sum((O - E)^2 / E); df = (rows - 1) x (columns - 1).
- AO3: Strengths: simple to compute and interpret, widely applicable to categorical data such as diagnosis categories, treatment uptake, survey response categories; does not assume normal distribution of the underlying variables since data are categorical.
- AO3: Strengths: provides a clear decision rule using critical values and p thresholds, useful for initial exploratory analysis of associations between categorical variables in applied settings.
- AO3: Limitations: assumes expected frequencies are sufficiently large; when expected cell counts are small (commonly any expected < 5) the approximation to the chi-squared distribution is poor and results may be invalid.
- AO3: Limitations: for small samples or 2x2 tables with low expected counts, exact tests such as Fisher's exact test or use of continuity corrections should be considered.
- AO3: Limitations: Chi-Squared indicates whether an association exists but not its strength or direction; measures of effect size (phi, Cramer's V) are needed to assess practical significance, especially in large samples where tiny effects can be statistically significant.
- AO3: Practical considerations: categorising continuous data to use Chi-Squared discards information and can be misleading; ensure categories are valid and meaningful in psychological contexts (e.g. diagnostic thresholds justified).
- AO3: Consider the role of sample size: large samples increase power but can make trivial associations significant; small samples may fail assumptions and lead to Type II errors if effect exists.
- AO3: Interpretation issues: significance does not imply causation, especially when data come from observational designs; researchers must consider confounding variables, sampling bias and independence of observations.
- AO3: Ethical and reporting considerations: report expected frequencies, df, X^2 value and p value, and effect size; be transparent about any continuity corrections or use of exact tests.
- AO3: Application to psychological examples: in a study of treatment uptake by diagnostic category, Chi-Squared can identify whether uptake differs by diagnosis, but follow-up with effect size and consideration of clinical importance is essential; discussion of generalisability if sample is clinic-based versus population-based.
- AO3: Balanced judgment: Chi-Squared is a valuable, accessible tool for categorical data but must be used with attention to assumptions, supplemented by effect size and cautious interpretation about causality and practical importance.
Question 12
- calculates the row totals, column totals and grand total correctly, e.g. leaflet total 60, no-leaflet total 60, contacted total 39, not contacted total 81, grand total 120 oeM1
- shows how to calculate an expected value using E = (row total x column total) / grand total and substitutes correct numbers for at least one cell oeM1
- gives the correct expected value for the contacted, leaflet cell as 19.5 and for the contacted, no-leaflet cell as 19.5 oeA1
- gives the correct expected value for the not-contacted, leaflet cell as 40.5 and for the not-contacted, no-leaflet cell as 40.5 oeA1
- substitutes O and E into the formula X2 = sum((O - E)2 / E) for each of the four cells oeM1
- calculates and states the numerical value for X2 to two decimal places as 3.08 awrt 3.08 oeA1
- calculates degrees of freedom correctly for the 2x2 table as df = (2 - 1) x (2 - 1) = 1 oeM1
- states the correct critical value at p ≤ 0.05 for df = 1 as 3.841 and compares it with the calculated X2 oeA1
- draws the correct conclusion that X2 = 3.08 < 3.841 so the result is not significant at p ≤ 0.05 and the null hypothesis is retained oeA1
- Answer: Totals: leaflet 60, no-leaflet 60, contacted 39, not contacted 81, grand total 120. Expected frequencies: contacted, leaflet = (60 x 39)/120 = 19.5; contacted, no-leaflet = 19.5; not contacted, leaflet = (60 x 81)/120 = 40.5; not contacted, no-leaflet = 40.5. Components: (24 - 19.5)2/19.5 = 1.0385, (36 - 40.5)2/40.5 = 0.5, (15 - 19.5)2/19.5 = 1.0385, (45 - 40.5)2/40.5 = 0.5. Sum X2 = 3.077 = 3.08 (2 dp). df = 1. Critical value at p ≤ 0.05 is 3.841. Since 3.08 < 3.841 the result is not significant at p ≤ 0.05; retain the null hypothesis, no clear association between leaflet and contacting counselling in this sample.
Mark your answers
Check answers without an account. Some answers can be checked automatically; use the mark scheme to mark the rest.