Chi-squared test

The chi-squared (χ²) test is a non-parametric statistical test used to determine if there is a significant difference between the expected frequency and the observed frequency in one or more categories. It is particularly useful for analyzing categorical data, where observations fall into distinct, non-overlapping groups. The core of the test lies in calculating the chi-squared statistic, which quantifies the discrepancy between the observed counts and the counts that would be expected if the null hypothesis were true.

The null hypothesis typically posits no association or no difference between the variables being studied. A larger chi-squared value indicates a greater divergence between observed and expected frequencies, suggesting that the null hypothesis may be false.

The chi-squared test was formally introduced by Karl Pearson in 1900 in his seminal paper 'On the Criterion of Goodness of Fit of a Distribution.' Pearson sought to develop a robust method for assessing how well observed data fit a theoretical distribution or for determining the independence of two categorical variables. Prior to his work, statistical inference for categorical data was less standardized. Pearson's contribution provided a unified framework, allowing researchers to rigorously test hypotheses about proportions and associations.

His criterion, often referred to as the 'goodness-of-fit' test, laid the groundwork for much of modern inferential statistics, particularly in the analysis of qualitative data.

The chi-squared test is a cornerstone of statistical analysis across numerous scientific disciplines due to its versatility and interpretability. In medicine, it's crucial for clinical trials to assess whether a new treatment is more effective than a control, or to identify associations between risk factors and diseases. In genetics, it helps determine if observed inheritance patterns conform to Mendelian ratios.

In social sciences, it's employed to explore relationships between demographic variables, opinions, and behaviors. Its ability to handle nominal and ordinal data makes it applicable where parametric tests, which assume continuous data and specific distributions, are inappropriate. The test empowers researchers to draw evidence-based conclusions and make informed decisions.

The chi-squared test operates through a systematic process. First, a null hypothesis (H₀) and an alternative hypothesis (H₁) are formulated. For a goodness-of-fit test, H₀ might state that the observed distribution matches a specific theoretical distribution, while H₁ states it does not.

For a test of independence, H₀ might state that two categorical variables are independent, and H₁ states they are dependent. Next, expected frequencies (E) are calculated for each category under the assumption that H₀ is true. The observed frequencies (O) are then collected from the data.

The chi-squared statistic is computed using the formula: χ² = Σ [(O - E)² / E]. This value is then compared to a critical value from the chi-squared distribution, determined by the degrees of freedom (related to the number of categories) and the chosen significance level (alpha, typically 0.05). If the calculated χ² exceeds the critical value, H₀ is rejected.

The chi-squared test encompasses two primary forms: the goodness-of-fit test and the test of independence. The goodness-of-fit test assesses whether a single categorical variable follows a hypothesized distribution (e.g., are the proportions of different colored marbles in a bag as claimed?). The test of independence examines whether there is a statistically significant association between two categorical variables (e.g., is there a relationship between gender and preference for a certain political party?).

It's important to note the assumptions: data must be counts or frequencies, categories must be mutually exclusive, and expected frequencies should generally be at least 5 in most cells to ensure the validity of the test's approximation to the chi-squared distribution. Violations, especially small expected counts, may necessitate alternative tests like Fisher's exact test.