When the results of correlation and regression are different, which one should I rely on more? For example, if the correlation of two variables is negative, but the direction is positive in regression or path analysis, how should I interpret the results?
Could you give us some data please, so we can have a look what's going on?
Correlation and regression are two statistical methods that are used to analyze the relationship between two variables. Correlation measures the strength and direction of the linear relationship between two variables, while regression is used to predict the value of one variable based on the value of another variable. When the results of correlation and regression are different, it is important to consider the context and the research question being addressed. In general, correlation is more appropriate for exploring the relationship between two variables, while regression is more appropriate for making predictions. If the correlation between two variables is negative but the direction is positive in regression or path analysis, it may indicate that there are other variables that are affecting the relationship between the two variables being analyzed. It is important to carefully interpret the results and consider additional analyses to better understand the relationship between the variables.
Lyudmil Antonov Thank you for your comprehensive answer. Much appreciated. One study had a multicollinearity issue between a predctor variable and another variable, which caused the issue I mentioned. I was wondering if you know there are any other statistical ways to identify why the inconsistent results emerge. Thanks!
Example see graph: R = 0.0525 because the data consist of 1 cluster of points with very little correlation and 1 distant point. The distant point doesn't contribute much to the calculation of R, but it does cause the parameters of the regression to be quite stable.
Inconsistent results can emerge due to a variety of reasons. Here are some statistical ways to identify potential issues:
Outliers: Outliers are extreme values that differ from most other data points in a dataset. They can have a big impact on your statistical analyses and skew the results of any hypothesis tests. There are four ways to identify outliers:
Sorting method Data visualization method Statistical tests (z scores) Interquartile range method Multicollinearity: This occurs when two or more predictor variables are highly correlated to each other, such that they do not provide unique or independent information in the regression model
The most common way to detect multicollinearity is by using the variance inflation factor (VIF), which measures the correlation and strength of correlation between the predictor variables in a regression model.
Inconsistency or Variation in Data: If the variation in data is above a certain limit, it might be necessary to discard that data. You could use robust measures like the Median Absolute Deviation or Interquartile range to compute a descriptive metric and perform outlier detection.
Measurement Errors: These can occur due to incorrect data entry, equipment malfunctions, or other measurement errors.
Sampling Errors: These occur when the sample is not representative of the population.
Model Specification Errors: These occur when the statistical model used does not correctly represent the underlying data-generating process.
- Badges
- Science topic
- Similar topics
- Psychology