They are aboth measurement of relationship, but the difference is Pearson gives coefficient of correlation and regression gives the contribution of IV for the correlation.

Another is, spearman is best for parametric data, while regression may be use with Non-parametric data.

Pearson is sometimes good with Nominal and Ordinal levels while Regression is good with Interval and ratio.

Pearson is good for continuous data while Regression is good for categorical data.

Dear Eekim,

If you want to assess relationship, so go for Correlation (Pearson/Spearman)

But if you want to assess effect, so go for regression.

Pearson for normally distributed data and Spearman for non normally distributed.

See these links:

https://www.graphpad.com/support/faq/what-is-the-difference-between-correlation-and-linear-regression/

https://support.minitab.com/en-us/minitab-express/1/help-and-how-to/modeling-statistics/regression/supporting-topics/basics/a-comparison-of-the-pearson-and-spearman-correlation-methods/

http://sites.utexas.edu/sos/guided/inferential/numeric/bivariate/cor/

The Pearson correlation measures the strength of the linear relationship between two continuous variables. The Spearman correlation coefficient is based on the ranked values for each variable rather than the raw data. Spearman correlation is often used to evaluate relationships involving ordinal variable

Single linear regression is used to show (study) how two variables (dependent and independent) linearly are related

thank you so much everyone! really helped alot!

Pearson correlation is to measure the relationship between variables, Spearman is to measure the relationship between ranks and simple linear regression is to assess the effect of one variable on another.

All the fellows have contributed well. Whenever we are to analyze the data, we ask ourselves whether we want to check whether variables are related or one is said to have impact on the other. If the case is of measuring strength of relation then we can use cross-tabulation (nominal or ordinal data), Spearman (ordinal data) and Pearson f(Continuous data).

But if the question is regarding one variable being the reason for the other then we need to use regression.

Eekim -

Why do you think that a sociodemographic characteristic could have anything to do with the accuracy of an ECG? You have to be careful not to find spurious relationships.

Since you said that you "...want to proceed with multiple linear regression," I think the most important area to consider then is regression. As Muhammad Mumtaz Khan said, we tend to think more in terms of causal relationships with regression (the x-variables may be "predictors" for the y-variable response), though that still may not necessarily be true.

I ran into an interesting example of a correlation that was not causal: I have had many skin cancers. A dermatologist once commented that people with hair so dark as mine rarely have such trouble with skin cancer. But what he was missing is that most people with such dark hair also have darker skin. That protects against skin cancer. Without that, the dark hair did not help me at all. Yet there is still a high negative correlation between dark hair and skin cancer.

For the multiple linear regression, you will note relationships with y by noting the regression coefficients (the beta estimates), but please also note that the x-variables are going to often have substantial impact on each other. They aren't exactly 'independent' of one another, but act as a group. In particular, when you have substantial collinearity among these variables, the sign of a coefficient can even change.

The following has examples and references which may be of interest:

"When Would Heteroscedasticity in Regression Occur?" 

Preprint, June 2021, J. Knaub, 

https://www.researchgate.net/publication/352134279_When_Would_Heteroscedasticity_in_Regression_Occur

For a regression, we think of the x-variables as fixed, but y is a random variable. It may be a combination of x-variables, the predicted-y part, but there is a random component, epsilon, but we can only measure e. Ideally, the variance of y given the predicted-y will become larger with larger predicted-y, which is heteroscedasticity in regression. Traditionally people have tried to transform to lose this feature, but I and others think it better to model the feature as it is, using weighted least squares, WLS, regression. (Note that for a time series regression with autocorrelation, GLS, generalized least squares regression is to deal with both heteroscedasticity and autocorrelation.)

Best wishes - Jim