Think about your theoretical or practical question
A lot of you worry about whether an analysis is significant or not. What I would like to see is that you engage with a theoretical or practical question. Identify a problem or intriguing question that can be answered. As consultants, you address problems. State your problem or question clearly. Find relevant literature or show practical examples that speak to the issue. Then look at your data to get an answer to your question or hypothesis. Is the hypothesis supported - good. Say what it means and where to go next (implications). If the results are not significant - this can be potentially be as meaningful. Why did you not see an effect? Is it a problem of how you tested it or does it tell you something meaningful about the phenomena?
A related set of questions goes along something like this: Do we need to do a mediation or is a regression fine? Shall we do an ANOVA or a regression? Again, my answer comes back to the theoretical or practical question. Choose the test that is appropriate for answering your question with the data at hand. Both mediation and regression can be meaningful, but they will give you slightly different answers (also note that mediation is based on regression ;) Decide whether an ANOVA or regression is better suited for the variables that you are looking at.
Think about the theoretical process
A few of you are exploring mediation and moderation processes. Fantastic! Mediation is about 'causal' processes. Think of a flow chart: variable A 'causes' variable B, which in turn then leads to changes in variable C. With this in mind, can satisfaction lead to more gender and then to more helping behaviour? Probably not in this physical universe (unless you have discovered a cunning option for turning happy people into a different gender). In cases like this, it often seems that moderation is the better option - for example, is satisfaction related to helping behaviour and does this relationship differ for males versus females?
Can satisfaction lead to more work performance which then leads to more health? Potentially, but you need to have a good rationale for it. Maybe a different ordering of your variables might make better sense in the context of your data.
How important are your results?
A few of you are exploring effect sizes (e.g., explained variance). This is great. Now the big question is what is a good effect size? Are 3.5% explained variance good or bad? Would somebody in your non-psych audience understand what explained variance means in the first place?
Translate these figures into something that is meaningful to a non-psych audience. For example, if you make $30,000 a year - would 3.5% more money be a good incentive for you to act or not? Alternatively, if you measured a variable with relatively little consequence (e.g., the happiness with your new garden chair) and you are able to explain 50% of that variability - is this important or meaningful for a manager? So think about the size of the effect and whether it important or meaningful. Baseline: What can you say about the effect and its importance for management?
Should I use a graph?
Yes - But! Think about what is on your graph and what a non-psych audience can take away from it. Putting a messy scatterplot with lots of dots and a funny line into your presentation may not be particularly meaningful. There are other ways of presenting correlations or regressions. Think of a flow chart or a path diagram - these can be interesting options for displaying relationships between variables. If you want to talk about the messiness of social sciences, a scatterplot can be good. But explain the key message that each graph or figure or picture is conveying. Colourful images just for the sake of it are not good communication!
Other issues? Post something on the discussion board or send me an email. If there are more common themes, I will update this post.
Overall, I love how engaged you are with your work. It is really cool and I sincerely look forward to seeing these presentations come alive :D