By graphing, you can see if the data is linear or not, basically you can analyze the structure of your data easily and you can then decide which model to use. If you have, say, one feature, you can graph the feature and its y value giving you a 2D graph (x-axis is the feature while y-axis is the y value). If you have 2 features then you will have a 3D graph. But what if you have 3 features or even more? Is it still possible to analyze your data using graph? I know it is impossible to graph if you have greater than 4 dimensions, but can you graph, like, 1 feature and the y value at a time? I thought I'll ask first for advice before performing experiments.

  • 3
    $\begingroup$ @Andre, I wouldn't recommend bubbles in 3D. Because the size will suggest visual perspective. $\endgroup$
    – ttnphns
    Jun 1, 2014 at 17:05

1 Answer 1


No graph can provide a complete analysis of the data, even for univariate or bivariate data. That is, a graph provides some insights but not all possible insights. For higher dimensional data, graphs still can provide insights, but they are relatively simple compared to the potential complexity of the data. Fortunately, even high dimensional data usually has relatively simple structure so low dimensional plots can still be very useful. Interactivity such as selection linking and filtering is especially useful for high dimensional data.

Some examples follow.

Histograms, looking at one variable per graph but linked by selection:

enter image description here

A scatter plot matrix, looking at every pairwise interaction:

enter image description here

Taking 2-D profiles, in this case 1 Y and 8 X variables. Each 2-D plot shows how Y varies by X when the other variables are fixed at the chosen values.

enter image description here

  • $\begingroup$ Is your first example on the scatter plot a binary classification? If you use scatter plot in a binary classification, won't you get just two horizontal line? $\endgroup$
    – hehe
    Jun 9, 2014 at 7:42
  • $\begingroup$ Yes, the chas variable is binary. The dots are not all in a line because random jitter has been applied to reduce overstriking. $\endgroup$
    – xan
    Jun 9, 2014 at 12:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.