Scatter Graphs

A scatter like this one has a very particular story to tell. Let us first consider those data points that are along the "top" of the "scatter". These particular data points are saying; "Even though 'A' was at maximum, for some reason we simultaneously had 'B' at several different values falling across the entire measurable range of 'B'." This is in fact true at all levels of 'A'. This is conclusive evidence that there is no real relationship at all between A and B. Neither variable is subject to common influence (because they do not "move" together), and certainly neither one "causes" the other.

This second example shows a scatter plot with more of a definite "shape" than the one above. This is an indication that there is indeed some kind of relationship between these two variables (though no graphing technique can ever tell you conclusively that one variable causes the other -- only that they are co-related, or "correlated" somehow). In this example, we see that by-and-large, as the value of 'A' decreases, the value of 'C' increases. As one decreases the other increases.  Because the general "slope" of the "data field" is pointed downward toward the right, we call this a "negative relationship" or sometimes an "inverse relationship".

Compared to the second graph, this graph at left shows a much more "tight scatter" or "grouping" of data points. This data field is definitely pointing "upward" toward the right. Because it is "tighter" we know that the relationship between 'A' and 'D' is linked much more strongly. Also, because it is definitely pointing "upward", we know that this is a "positive relationship". This means that as the value of one variable increases, so too does the other variable increase, by a largely proportionate degree (which is why this is a "tight" scatter).

In your Quality Improvement work you are likely never going to see a "direct relationship". This is when a movement of one variable produces an even and constant movement of the other variable - across the entire measurable range. In a case like this, all of the data points fall on top of a single line. Because they all fall on a line, this is sometimes called a "linear relationship", though the term really means that the line is straight. Sometimes the data line-up on a single line, but that line is not straight, it is curved. This is known as a "curvilinear relationship", and is typical of accelerating phenomena, such as "diminishing returns", "accelerating loss", and even "collapse" or "seizure" (or "critical mass" at the point of inflection - not depicted in these diagrams).