- Visual clarity, which is the idea that when making a figure, one should strive to make the data stand out and avoid superfluity in your graph. This means things like ensuring data points aren't buried by trend lines, axes, labels, or other forms of clutter.
- A clear understanding of a graph's purpose, by way of putting the conclusions from which you want the reader to draw in graphical form and making captions comprehensive and informative. A good caption should describe everything that is graphed in the figure, draw attention to the important points, and describe the conclusions to be drawn from the graph.
- Aspect ratios and the idea of banking to 45 degrees. This is the idea that humans are best able to estimate rates of change in graphical data when the average slope from point to point is 45 degrees. Making your figure dimensions to bring the plotted data close to this is probably a good idea.
- Scales, and the appropriate choices of tick marks and scales to compare data between graphs.
Some good resources for general scientific figure creation and examples, both good and bad, include:
along with classic books like Tufte's The Visual Display of Quantitative Information and Cleveland's The Elements of Graphing Data.