The Chi-Square Test in R

In the field of statistics, the Chi-Square Test is one of the most commonly used tools for non-parametric hypothesis testing involving categorical data. It is primarily used to determine whether there is a significant association between two or more categorical variables, or whether the difference between observed frequencies and expected frequencies is caused by random error. For beginners in R, mastering this method not only helps solve practical problems but is also a crucial step in understanding statistical inference thinking.

The basic logic of the Chi-Square Test is built upon the assumption of "independence." If two categorical variables are independent, then the expected frequencies calculated from the marginal distributions should be very close to the actual observed frequencies; conversely, if there is a significant association between them, large deviations will occur between the observed and expected frequencies. The built-in chisq.test() function in R can automatically complete the complex calculation process, including constructing contingency tables, calculating the Chi-Square statistic ($\chi^2$), degrees of freedom (df), and the corresponding P-value, thereby helping researchers make objective decisions.
To perform the simplest Chi-Square Test in R usually requires only two steps: first, construct a matrix or data frame containing the observed frequencies, and then call the chisq.test() function. This function will automatically determine the dimensionality of the data and output detailed statistical results.

Suppose we have survey data regarding "Gender" and "Purchase Behavior," aiming to analyze whether there is a significant difference in preference for electronic products between males and females. We can create a $2 \times 2$ contingency table:

# Create observed frequency matrix
data_matrix <- matrix(c(50, 30, 45, 25), 
                      nrow = 2, 
                      byrow = TRUE, 
                      dimnames = list(Gender = c("Male", "Female"), 
                                      Purchase = c("Yes", "No")))

# Perform Chi-Square independence test
result <- chisq.test(data_matrix)

# Output results
print(result)

After running the code above, R will return an object containing the Chi-Square statistic, degrees of freedom, and P-value. If the P-value is less than the preset significance level (usually 0.05), we reject the null hypothesis and conclude that there is a significant association between gender and purchase behavior; otherwise, we consider them to be independent. It is important to note that when the contingency table contains expected frequencies that are too small (typically less than 5), the results of the Chi-Square Test may no longer be accurate. In such cases, Fisher's Exact Test might be required. R also supports this via the fisher.test() function.

Handling Multivariate Data and Goodness-of-Fit Tests

In addition to the most common independence test, the Chi-Square Test can also be used for Goodness-of-Fit Tests. In such scenarios, we are not concerned with the relationship between variables but rather focus on whether the distribution of observed data conforms to a specific theoretical distribution. For example, rolling a die six times and observing whether the counts of each face appear uniformly distributed.

In R, the chisq.test() function is very flexible; it can directly accept numeric vectors as input and specify the expected probability distribution. The following is a simple example used to test whether a set of data follows a normal distribution (simplified here to a uniform distribution for demonstration):

# Generate simulated data: assuming rolling a die 120 times
observed_counts <- c(18, 20, 19, 21, 17, 25) # Counts for each face

# Define expected frequencies (uniform distribution, total/6)
expected_counts <- rep(20, 6)

# Perform goodness-of-fit test
fit_test <- chisq.test(observed_counts, p = c(1/6, 1/6, 1/6, 1/6, 1/6, 1/6))
print(fit_test)

In this example, we manually specified the expected probability for each category as $1/6$. If the P-value is significant, it indicates that the data distribution differs significantly from the theoretical uniform distribution, suggesting the die may be uneven or the experimental process has bias. This application demonstrates the wide applicability of the Chi-Square Test in quality control and data analysis.

Case Study: Educational Background and Career Choice

To better understand the application of the Chi-Square Test in practical research, let us construct a slightly more complex case. A school surveyed 200 students, recording their "Educational Level" (Bachelor's, Master's) and "Preferred Career Direction" (Technical, Management). We want to know if higher education significantly influences students' career choice tendencies.

First, organize the data into contingency table format:

# Construct contingency table
education_job <- matrix(c(30, 25, 
                          15, 40), 
                        nrow = 2, 
                        dimnames = list(Education = c("Bachelor", "Master"), 
                                        Job_Type = c("Tech", "Mgmt")))

# Perform test
chi_sq_result <- chisq.test(education_job)

# View summary of results
summary(chi_sq_result)

Using the summary() function, we can clearly see the value of the Chi-Square statistic and its accompanying P-value. Assuming the result shows a P-value of 0.012, this means that at a 98.8% confidence level, we have reason to believe there is a significant association between educational level and career choice. Specifically, the data might show that the proportion of Master's students in the technical field is significantly higher than that of Bachelor's students, or vice versa.

In practical operations, researchers often not only focus on the P-value but also examine effect size (Effect Size). Although chisq.test() does not directly output effect size metrics like Cramér's V by default, combining it with matrix operations can further quantify the strength of the association, which is crucial for writing academic papers or conducting deep data analysis. By systematically applying the Chi-Square Test, we can extract valuable statistical patterns from chaotic categorical data.