Trends combine your facts and stats to reveal more sophisticated patterns. You can explore the difference within a group (e.g., change in knowledge before and after an intervention) or between groups (e.g., level of satisfaction between two different programs).
Find the Patterns: Trends
To answer your strategic questions and gain a deeper understanding of your program’s impact, you will need to take your raw data and transform it using a combination of basic arithmetic and descriptive statistics to reveal patterns – both those that you expect to see and those that may be surprising.
Trends using facts: Take a collection of facts, like the count of program participants, and put those counts from several programs together in one table or chart.
Trends using stats: Take a collection of stats, like a collection of means or distributions, and look at them together in a table or chart.
- Trends using averages: Calculate the average level of satisfaction with each of the program sessions, where 1 is very unsatisfied and 5 is very satisfied. Put those averages into a table and if you like, also create a bar chart to compare them.
- Trends using distribution: You can look at distribution trends to see changes, like if satisfaction with a program changes from year to year. For each year of the program, calculate the percentage of people who responded that the program exceeded, met or did not meet expectations. Create a table with those percentages, and a stacked bar graph to look at the difference visually.
One common method of looking for trends is by examining differences between groups: For example, are teens who attend a private high school more likely to choose a service gap year than teens who attend a public high school?
Looking at change over time within a group can again be accomplished by analyzing your data for both central tendency and distribution.
Central tendency: Compare average scores on pre- and post-tests.
|Test 1||Test 2|
Did your program increase knowledge about gap year service programs, and is that change measurable? Let us assume that you gave a pre- and post-test to your program participants testing how much they know about what gap year service programs are out there for them to consider. Looking at this chart, the average score is higher on the second test than on the first, meaning knowledge increased after the program.
Distribution: You can also look at your pre- and post-test scores using distribution statistics. Here are all of your test scores, divided into pre- and post-test scores, in a scatter plot.
It may be a bit difficult to draw anything definitive from this scatter plot, although it does appear that there are more blue dots (pre-test scores) in the lower half of the chart and more green dots (post-test scores) in the upper half of the chart. This indicates that scores on post-tests were higher than on pre-tests.
It may be easier to see this result by looking at the distribution in a table.
This table shows that more people scored 80 or above on the post-test.
|Score||Test 1||Test 2|
You could even further summarize the distribution depending on the goal of the program. For example, if the goal was to increase knowledge above a score of 80, you could use this chart:
|Test 1||Test 2||Change|
Advanced: You could run an advanced statistical test on this data to calculate whether the difference in test scores is statistically significant,meaning it is statistically likely that the change you are seeing is not the result of chance, and the change in scores can be (at least in part) attributed to the class. The statistical test is called a Paired t-test, and you can find out how to run the test in Excel using this tutorial.
Looking at change over time between groups can be as easy as using simple arithmetic (e.g., counting the number of program participants).
Pattern spotted: Program participation is declining
You can also look at change over time between groups using central tendency and distribution. For example, we could explore changes in knowledge between many different groups, instead of just one group as we explored above.
Pattern spotted: Year 2 saw a more significant increase between pre- and post-test scores than Year 1.
Sometimes you will discover relationships between variables (e.g., individuals age 20-25 are more likely to be highly satisfied with your program than individuals older than 25).
How do you look for these kinds of relationships?
- Make a list of demographic and other characteristics you think might be interesting to explore
- Recalculate your facts and stats by different groups (e.g., men and women) to see what the data shows you. You can do this in tables or charts.
Let's look at an example of test scores (variable 1) vs. gender (variable 2). Again, it may be a bit difficult to draw anything definitive from a scatter plot, although it does appear that women scored higher than men. This is easier to discern in a table, where we can see the average score for women was higher than for men.
Advanced analysis: You could run a statistical test that measures the strength of the relationship, or the correlation, between test scores and gender.
If you have large data sets and the organizational capacity for more complex analysis, you can begin to look at regression and prediction to help you tailor programming to meet your audience’s needs. For example, finding the right mix of experiences for a college student that will maximize the chance they choose to work in the nonprofit sector. Here is a tutorial on running regression and multiple regression analysis in Excel. Be aware that Excel has some shortcomings with more complex analysis because it was not originally developed to support it. Instead, it has been adapted for more complex analysis with add-ons.