Interpreting results and presenting findings Intermediate Food Security Analysis Training Rome, July 2010
Overview • Determining the question you want to answer • Using your analysis plan • Interpreting results from SPSS • Visualizing findings • Writing-up your analysis
Determining the question you want to answer • The key questions we typically try to answer in a CFSVA are: • Who are the food insecure people? • How many are food insecure? • Where do they live? • Why are they food insecure? • For each question, we must first think about the analysis we need
What is an analysis plan? • The link with the conceptual framework that sets out your hypotheses • A table detailing data to be collected and how those data will be analyzed • A guide to the analytical process
Think back to your analysis plan • Who are the food insecure people? • Cross-tabulate various demographic indicators with food consumption groups • Sex of household head • Dependency issues • Education • Etc. • Verify differences are significant using hypothesis testing • Is the sex of the household head a significant factor different between the food secure and the food insecure? • Are households with a high percentage of dependents significantly more food insecure? • Does education significantly affect food security?
Thinking about an analysis plan • How many are food insecure? • Run a frequency on food security groups • Where do they live? • Cross-tabulate food consumption groups by strata • Urban / rural • Agro-ecological / livelihood zones (if available) • Administrative zones (governorates, provinces, districts, etc.) • Always verify differences are significant using hypothesis testing
Thinking about an analysis plan • Why are they food insecure? (a bit out of scope for this training, but good to think about) • Keep the conceptual framework for food security analysis in mind and explore the dataset using the tools you have available to you • Run hypothesis tests on the various data you have. For example: • Exposure to shocks • Coping strategies index • Ability to cope with shocks • Wealth • Access to credit • Types of livelihoods • Access to markets • Etc. • Use regression analysis (in the next training!)
Presenting results: a few pointers • A good graph must convey statistical information quickly and efficiently • The minimum ink principle • Avoid images with 3-D effects or fancy shading. Use the minimum amount of ink to get your point across. • The small table principle • A small table is better than a large graph. If you graph contains 20 data points or less, use a table of numbers instead. • The rule of seven • If a table has seven or more rows or columns, it probably has more information that can be easily interpreted • The fault of default principle • Don't rely on the default options when creating graphs. Try multiple versions until you get the right information presented
Presenting results: using color • Danger in the use of color • Color should be avoided for ordinal data • Shades work better with ordinal data • Bright colors can lead to optical illusions • For example, areas in bright red sometimes appear larger than areas in bright green • Certain color combinations are difficult to distinguish • Blue against a black background • Yellow against a white background • More than 8% of all males and more than 1% of all females are colorblind • A red-green deficiency is most common • Color is often culturally biased
A few points about tables • Show only two significant digits at most • If possible, sort rows with the largest numbers at the top • If you’re showing the same rows (strata of analysis) repeatedly, you should consider being consistent in the order of the rows • Use a table anytime you have 20 or fewer numbers.
Example of an area chart – FCS/ Food group composition Consumption frequency Food consumption score
Example of a segmented bar graph – food consumption groups by marital status
Interpreting results from SPSS • Once you’ve created an analysis plan you can start your work in SPSS • Each output of SPSS has a lot of information. Understanding these outputs is critical. • What do the ANOVA results below tell you about share of food expenditure between urban and rural populations? • What would you share in your findings?
Presenting your results • Never use SPSS outputs for sharing your results! • In this case, a very simple table can illustrate that rural populations spend a larger share on food than urban populations • In the text that describes the table, we can note the statistical significance (depending on our audience) Table 1 – Average share of food expenditure by urban / rural The results from the survey showed that rural populations significantly (p<0.05) spent a larger share on food than urban populations, 47.5% as compared to 39.0% respectively.
Sharing results • Consider the table below. Does it clearly illustrate any information?
Sharing results • Generally speaking, ‘the rule of seven’ should be applied during report writing. If a table has more than six rows or columns, it probably has more information that can be easily interpreted. Consider creating a graphic or creating a table with just the pertinent information
Sharing results • Simply sorting data can make a graph much easier to read and can quickly highlight the point you are illustrating • What is missing from this table? Figure 1: Education level of household head by Governorate
Writing up your results • Always remember the question you are trying to answer when writing. • Have a solidly defined report structure prepared before you write. The analysis plan can help you with this • Don’t make assumptions that you cannot backup • Think of the results as telling a story. You need to build your findings over the course of the story and transition from section to section as fluidly as possible. Use the conceptual framework to guide you. • Use visual aids to highlight your points, but don’t rely on them to do all the work. Make sure you have meaningful titles • Get your colleagues to review your work!