How to evaluate quality of correlation?

In summary, the evaluation of correlation quality involves looking at the R-values, which range from -1 to 1, and focusing on values of at least 0.7. However, the interpretation of these values also depends on the nature of the data and its intended purpose. Non-linear trends and other factors should also be considered when evaluating correlation.
  • #1
simpleton
58
0
Hi,

I would like to know, how do I evaluate the quality of correlation?

Specifically, I have a set of N datapoints, each represented by k features. I want to know how the k features correlate with each other, and therefore, I created a k by k correlation matrix. I am using R-value, so the values range from -1 to 1. I was told to only look at R-values of at least 0.7, because anything lower than that does not mean much. However, I was wondering, how do I evaluate a correlation value of say, 0.85? If a feature x has a R-value of 0.7 with y and a R-value of 0.9 with z, what quantitative conclusions can I draw and what kind of statements can I make? For example, will I be able to say how much better the correlation between (x,z) is compared to (x,y)?

Thank you very much!
 
Physics news on Phys.org
  • #2
Hey simpleton.

The answer to your question is highly subjective.

You need to not only take into account the correlation value, but the nature of the data which include the graphical nature as well as the domain in which the sample is represented.

If there is a non-linear trend with what looks to be a seemingly low-residual, then using the linear correlation method will not really emphasize the true relationship between the variables.

The above points vary depending on what kind of data it is, where it came from, what its used for, and what questions its trying to answer.
 

What is correlation and why is it important to evaluate its quality?

Correlation is a statistical measure that shows the strength and direction of the relationship between two variables. It helps us understand how closely related two variables are and if changes in one variable result in changes in the other. It is important to evaluate the quality of correlation because it allows us to determine the credibility and usefulness of our findings.

What are the different types of correlation and how do they differ in terms of quality evaluation?

There are three types of correlation: positive, negative, and zero. Positive correlation indicates that as one variable increases, the other variable also increases. Negative correlation indicates that as one variable increases, the other variable decreases. Zero correlation indicates that there is no relationship between the two variables. The quality of correlation is evaluated by calculating the correlation coefficient, which ranges from -1 to 1. The closer the coefficient is to -1 or 1, the stronger the correlation, while a coefficient close to 0 indicates a weak or no correlation.

What are the limitations of using correlation to evaluate relationships between variables?

Correlation does not imply causation, which means that just because two variables are highly correlated, it does not mean that one variable causes the other to change. There may be other underlying factors that are responsible for the observed correlation. Additionally, correlation cannot determine the direction of the relationship between two variables, as it only measures the strength and not the cause-effect relationship.

What are the potential sources of error when evaluating the quality of correlation?

One potential source of error is using a small or biased sample size, which can lead to inaccurate results. Another source of error is omitting relevant variables that could impact the relationship between the two variables being studied. It is also important to consider potential outliers or influential data points that could skew the results.

How can we improve the quality of correlation evaluation?

To improve the quality of correlation evaluation, it is important to use a larger and more diverse sample size to ensure the results are representative of the population. Including relevant variables and controlling for potential confounding factors can also improve the accuracy of the correlation. It is also important to use appropriate statistical tests and techniques to analyze the data and interpret the results correctly.

Similar threads

I Variable Normalization for different variable ranges....
    • Set Theory, Logic, Probability, Statistics
Replies
3
Views
1K
I Vector visualization of multicollinearity
    • Set Theory, Logic, Probability, Statistics
Replies
1
Views
615
I Linear regression and random variables
    • Set Theory, Logic, Probability, Statistics
Replies
30
Views
2K
B Does a correlation coefficient represent probability?
    • Set Theory, Logic, Probability, Statistics
Replies
4
Views
5K
0.7 correlation but almost 0 p-value, how to interpret?
    • Set Theory, Logic, Probability, Statistics
Replies
4
Views
2K
MHB Calculating Heteroscedasticity: Is Correlation Between Variables Strong/Weak?
    • Set Theory, Logic, Probability, Statistics
Replies
9
Views
2K
I Weird covariance matrix for linear fit
    • Set Theory, Logic, Probability, Statistics
Replies
1
Views
837
Bound correlation coefficient for three random variables
    • Precalculus Mathematics Homework Help
Replies
3
Views
2K
I Correlation vs causation in biology/biochemistry
    • Set Theory, Logic, Probability, Statistics
Replies
21
Views
3K
CDF of correlated mixed random variables
    • Set Theory, Logic, Probability, Statistics
Replies
9
Views
2K

Hot Threads

Recent Insights

Back
Top