The Results section provides a summary of all data collected in your study, along with the main trends which can be observed in the data, but includes no opinions and interpretation (these should be reserved for the Discussion). Three formats are available for presenting information in your Results — text in paragraph form, tables, and figures (usually graphs). Which format you adopt for a specific dataset will depend on what type of data you are summarizing, with one exception. Every Results section should include at least a few paragraphs of text describing the main trends, and pointing readers to the appropriate tables and graphs for further details.
What are trends?
Trends are the general patterns we can see in our data. A difference in means or distributions, or a correlation between variables, is a trend. Trends may be apparent by looking at the data, but statistical tests provide a more objective way of assessing whether the patterns are real. Therefore, descriptions of trends are typically paired with the results of the statistical test performed. For example: “We found a significant difference in mean body weight between rats fed Purina Rat Chow and rats fed potato chips (two-tailed t-test with equal variance, df = 94, t = 4.51, p < 0.01)." Here the trend is the difference in mean body weight, and we believe the difference is significant (i.e. not simply due to random chance) because our t-test tells us our results are highly unlikely given our null hypothesis.
When we hypothesize that a trend exists, but don't find it, we need to note this too. For example, "Rats fed wieners did not differ significantly in body weight from rats fed lard (two-tailed t-test with equal variance, df = 74, t = 0.65, p > 0.05).
How should I summarize my data?
How you summarize your data will depend on the type of data you are presenting. Simple results can be clearly and conveniently presented in text form. More complex data may be easier to read when organized into a table. Graphs are a powerful way of showing trends because humans tend to be visual thinkers. We see patterns more easily than we can deduce them abstractly.
Some types of data lend themselves to descriptive statistics. For example, if you are comparing two means you would naturally want to include the means, along with some measure of variation in each treatment group, and the sample sizes. Some types of data cannot easily be summarized, and can only be presented in tabular (table) or graphical form. An example would be a comparison of two distributions. If you counted 52 women and 34 men in the university gym over a 24 hour period, you can’t estimate central tendency or variation. You just have to present the numbers. And correlation data are even harder to summarize. All you can do with the data is graph them (fortunately, graphs are the most powerful way to display trends).
You should almost never present raw data in your paper. Even a small data set will be impossible for the reader to absorb and understand without some work on your part to summarize, organize or graph to show the important trends.
Tables:
Tables are numbers presented in rows and columns. Tables should include an informative title at the top. The title should provide enough information that the reader can understand the table without referring back to the text. Tables should be numbered based on the order they are referred to in the text. Tables may be interspersed in text, or gathered at the end of the paper. If they are included in the body of the paper, they should never appear before they are referred to in the text.
Figures:
Figures include all visual elements included in your paper. Figures should include an informative caption at the bottom. The caption should provide enough information that the reader can understand the figure without referring back to the text (does this sound familiar?). Figures should be numbered based on the order they are referred to in the text. Figures may be interspersed in text, or gathered at the end of the paper. (Tables and figures should be gathered in separate collections). If they are included in the body of the paper, they should never appear before they are referred to in the text.
By far the most common type of figure in the Results section is the graph. Many types of graph are available (e.g. bar, scatterplot, pie) and the best type to use will depend on the data you are summarizing and the trends you are attempting to illustrate. Try to draw the reader’s eye to the comparisons or correlations you describe in the text. Your goal should be to present as much information as possible in the graph, but make it clear and intuitive. A well thought out and carefully prepared graph can not only help your reader to see patterns in the data, it can help you as well. Visual displays may show patterns that even statistics have not revealed (Tufte 1983).
General tips for graphs:
Most graphs have an X (horizontal) and Y (vertical) axis and these should be clearly labeled, including units if appropriate. When presenting results of an experiment, it is customary to graph the dependent variable on the Y-axis and the independent variable on the X-axis. When preparing a scatterplot of correlation data, either variable can go on either axis.
Caption should interpret data points if necessary. For example, if each point represents the mean of five measurements, this should be indicated. If error bars are used (and they should be included whenever data points are means) they should also be explained. (Are they variance, standard deviation, or standard error?)
If you are plotting points from several datasets on the same graph, make sure you include a legend to show the reader which point belongs to which dataset. The legend can be included as part of the caption if this is easier (this may not be acceptable to all instructors).
Use a scale which is appropriate for your data. Use all the space available on your graph, rather than having all your points cramped up in one corner.
Don’t add a title to your graph. All necessary information should be in the caption at the bottom.
Don’t connect data points with a line, unless that line means something. A simple “connect the dots” jagged line rarely makes things clearer and usually has no meaning with respect to the data. Including the regression line is appropriate if regression analysis has been performed on the data, but is not appropriate if the data were analyzed by a simple test of correlation.
Many journals and some instructors do not accept colour graphs. Check to see what is appropriate for your paper.
Many computer graphing programs are available and can make preparing a professional-looking graph much simpler. But they are not a substitute for knowing how to design an effective graph. The default settings on these programs will almost never produce an appropriate scientific graph. Learn how to use the program to produce the graph you want, rather than just copying a standard template supplied by the programmers.
Excellent references are available to help you learn to design great-looking graphs (e.g. Tufte 1983).
Choosing a Graph type:
The proper format for your graph (line graph, bar graph, etc.) will depend on the type of data you are dealing with, and the trend you are trying to display. Choosing an inappropriate format will result in a graph which is a best difficult to interpret, and at worst totally incomprehensible.
First ask yourself what type of data you are displaying on each axis of your graph. With nominal data on your x-axis and ordinal or ratio/interval data on your y-axis, you would likely want to use something like a bar graph (discrete categories on your x-axis, and continuous data on the y-axis.) With ratio/interval data on both axes, a scatterplot is likely more appropriate.
Also ask yourself what trend you want the graph to display to the reader. If you are looking for a difference between two or more treatments you will probably want something like a bar graph (discrete categories on your x-axis and continuous data on the y-axis). If you are looking for a correlation, you will probably need to have continuous data on both axes, and use a scatterplot (with or without a regression line).