It really depends a lot on the context. The Cohen scale, for example, puts a correlation of 0.5 right on the cusp of "medium/modest" (0.30-0.49) and "large/strong" (0.50-1.00). However these cutoff criteria are largely arbitrary and shouldn't be applied too strictly.
A correlation of 0.5 might be regarded as strong in some social science situations (e.g. where the measures are based on 5-point Likert scales) or weak in physical science situations where instrumentation can be extremely precise.
A perfect positive correlation would be exactly 1; 1.00 means "0.995 or higher", which is quite strong indeed.
An undefined correlation is one in which the data would not plot with points making a vertical line.
Absolutely not. The simplest way to demonstrate this is to consider a measure of agreement - disagreement. If we scored it so that "strongly agree" is 5 and "strongly disagree" is 1, we would get one value of the correlation. If we reverse-scored it, we would get exactly the same value, but with the opposite sign. The strength of the correlation is the same, but the direction of the relation has switched. Another consideration is the fact that the actual strength of the correlation is based on the square of its value. 0.20 squared is 0.04; 0.40 squared is 0.16. A correlation of 0.40 is four times as strong as a correlation of 0.20. But when you square something, you automatically lose the sign. The square of a negative number is positive. So by definition, correlations of the same size but different signs are equal in strength.
Yes. * A positive correlation is when the dependant variable increases as the independent one does. * A negative correlation is when the dependant variable decreases as the independent one increases. * Perfect correlation is when all the points lie along a straight line; no correlation is when the points lie all over the place. In calculating the correlation coefficient it can have a value between -1 and 1, with 0 indication no correlation and values between 0 and ±1 showing a greater correlation until ±1 which is perfect correlation. Moderate correlation would be one of these intermediate values, eg ±0.5, which shows the points are moderately related.
If you remove certain data points from a dataset, the correlation coefficient may be affected depending on the nature of the relationship between the removed data points and the remaining data points. If the removed data points have a strong relationship with the remaining data, the correlation coefficient may change significantly. However, if the removed data points have a weak or no relationship with the remaining data, the impact on the correlation coefficient may be minimal.
"Strong" is very much a subjective term. Not only that, but it depends on expectations. In economics I would consider 70% to be a strong correlation, but for physics I would want more than 95% before I called the correlation strong!
No, The correlation can not be over 1. An example of a strong correlation would be .99
A perfect positive correlation would be exactly 1; 1.00 means "0.995 or higher", which is quite strong indeed.
If the form is nonlinear (like if the data is in the shape of a parabola) then there could be a strong association and weak correlation.
The graph follows a very strong downward trend. Would have helped if you specified which correlation coefficient; there are different types.
Pearson's Product Moment Correlation Coefficient indicates how strong the relationship between variables is. A PMCC of zero or very close would mean a very weak correlation. A PMCC of around 1 means a strong correlation.
Anything modest and conservative.
they are the same. +1.00 and -1.00 are the strongest correlations. If you have +.92 and -.92 then that's a strong correlation but if you have -.15 and +.15 then that would be a weak correlation. There for + 1 or - 1 makes no difference
1 would be the strongest possible. 0.353 seems to be on the weak side. Above 0.4 or 0.45 may be strong enough.
An error! Correlation must be between -1 and 1.
I would gladly tell you what a wonderful person I am but alas, I am too modest. But I am also modest for saying that.
There would be no definite correlation. It would just be a random correlation that would be all over the graph because there is no trend in hair color and weight. Your weight doesn't determine your hair color.