Graphs are a visual representation of your data and as such should be easy to understand but more importantly, to interpret.
Basic rules:
• Every graph should have a underlined title that will allow the person reading your work to understand exactly what it will show without seeing the graph and should be the same title as the graph it is based on.
• Each axis has to be clearly labelled and have units.
• Remove all gridlines, it clutters the graph.
• The numbers on the x axis should be on tick marks, not between.
• There should be no fill, ie no background colour.
• The graph should be of a decent size, half a page is a good size.
• Do not have a legend unless you have more than one data series.
• If you are going to use coloured lines, make sure you print in colour.
• Add error bars.
• Add a trendline.
The graph you chose depends on the dependent variable, if it is categories (discontinuous) then you use a bar graph but if it is linear data (continuous) use a scatter graph.
Always add a trendline when using a scatter graph.
Make sure that it is clear to which points the trendline applies if you have more than one data series. Use the same colour for the line as you do for the trendline as the software often makes all trendlines black by default and it can get confusing.
Error bars can be for both the x and the y axis or just one of them depending on what you wish to show. An easy rule to follow is that you should show error bars for the dependent variables as this is where you are aiming to show range of possible error but if you have manipulated the independent variable (making different concentrations of saline solution for example), then you could also include error bars. Error bars can be based on uncertainty, standard error of the mean and standard deviation. If based on uncertainties, it can either be as a percentage or as a quantitative value.
You can plot more than one data series on a graph but only if you need to compare them; if not, do separate graphs.
If you have more than one data series in your graph, make sure you have a key that indicates the difference between each data series.
