Skip to content
TheChatArchiveThe Chat Archive
Mistralmixtral-8x7b-instruct-v0.1· Mistral AIHugging Face Arena Human Preference 55K

When do I use Matthew corealation coeficient and when Pearson correlation coeficient?

From a Chatbot Arena battleThe same person sent this prompt to two anonymous models at once. Compare it with Metallama-2-70b-chat’s answer.
  1. User

    When do I use Matthew corealation coeficient and when Pearson correlation coeficient?

  2. Mistral Assistant

    Both the Matthew correlation coefficient (MCC) and Pearson correlation coefficient (PCC) are measures of the strength and direction of the linear relationship between two variables. However, they are used in different contexts and have different assumptions.

    The Pearson correlation coefficient is a measure of the linear correlation between two continuous variables and assumes that the relationship between the variables is approximately normal. It ranges from -1 to 1, where -1 indicates a perfect negative linear relationship, 1 indicates a perfect positive linear relationship, and 0 indicates no linear relationship. The PCC is sensitive to outliers and assumes that the relationship between the variables is linear.

    On the other hand, the Matthew correlation coefficient is a measure of the correlation between two binary variables, i.e., variables that can take only two values, such as 0 and 1. It is often used in machine learning and bioinformatics to evaluate the performance of binary classifiers. The MCC ranges from -1 to 1, where -1 indicates total disagreement between the predicted and actual values, 1 indicates perfect agreement, and 0 indicates no better than random guessing. Unlike the PCC, the MCC is less sensitive to imbalanced data and can handle situations where there are more negative or positive samples.

    Therefore, you should use the Matthew correlation coefficient when you have binary variables and want to evaluate the performance of a binary classifier, and use the Pearson correlation coefficient when you have continuous variables and want to measure the strength and direction of the linear relationship between them.

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.