Once you know what graph type you want to use, there are still decisions to be made! Let me show you your options...
TABLE
Can reinforce trust with the user because you can be precise, and pull out an exact value that you want to refer to.
Be sure to consider: row banding, font colour, header titles, data field alignment, units and correct orientation (pivoting).
BAR
TRADITIONAL: Familiar to people. A fan favourite.
HISTOGRAM: Superb at showing distribution patterns.
STACKED BARS: Can be effective for % totals, when dealing with two categories. Otherwise I will avoid.
WATERFALL: Effectively a broken up, clustered bar graph
(CLUSTERED BAR GRAPH: Handle with care)
LINE
TRADITIONAL: Shows a relationship between variables with ease
CYCLE PLOT: A line graph that shows two trends at the same time, e.g. sales by months and year
SLOPE CHART: Fantastic for a specific look at the data between two points, especially changes of a measure between two date points
SPARKLINES: Great for finding context. Don't have to be the star of the show, can hide in a tooltip
(AREA: not a great choice. Keep scrolling)
SCATTER
TRADITIONAL: Shows the strength of a correlation between two measures brilliantly. And obviously.
QUADRANT: Very effective for subsequent analysis, with very little input. You just put two average lines in, one for each axis, and compare the 4 quadrants. Sales and profit are great measures to compare in this format.
MAP
TRADITIONAL: essentially a scatter graph. It is just a latitude and longditude coordinates, which are two measures, over a background. Maps already come with a lot of contextual knoweledge from the user. This is fantastic, but with great power comes great responsibility. Do not bias the user. Or yourself.
SHAPE MAP: Will show points on a map. Great to plot individual records. Maintains granularity. Not ideal for points covering huge areas.
CHLOROPLETH MAP: Essentially, it is shading by section. The section is usually a location sub-section, such as region, state or post code. Great for aggregated data, such as averages.
TILE MAP: Normalizes the size of different areas on the map which makes comparison easier. However, you need to watch out on the relationship the tiles have with each other, and check whether that is the relationship in real life. Europe fits nicely together on the map, but do the states of the US?
DENSITY MAP: Like a heat map and shape map (individual instances) combined. Great at showing 'hot spots'
PART-TO-WHOLE
PIE CHART/ DONUT: To be used with great care. Potentially with binary responces, like 'yes' or 'no'. Definitely do not use for more than 3 catergories.
TREE MAP: Appear initally confusing. But once you get your head around them, they can be very effective, especially at showing proportionality between categories, and dominance within a category e.g. manufacturers of a product.