I asked the same respondents about how much luck, their own efforts and connections they need to achieve Goal # 1, # 2, # 3, # 4. (first group). The second group had the same questions and the same goals, but they were in different conditions.
I analyze my data as a mixed plan, BUT the peculiarity of my design is that each respondent answered in the format: % luck +% efforts +% connections = 100%. Other answers were not accepted by the system, the amount always was 100. (a+b+c=100)
In SPSS I have 3 different variables: Luck_Goal1, Effort_Goal1, Connection_Goal1. And so on for goals 2, 3, 4 … But these aren't exactly three different variables, are they? It's like one variable, like a triad, and can be compared to other triads ...
My question is, how to combine this three-part variable into one whole triad variable, and how will intragroup and intergroup analysis of triads be done?
Many thanks,
Oleksandra
It seems if a+b+c=100 then c linearly depends on a and b, so you need only a and b for regression.
Andrey Davydenko is right. This constraint is analogous Designed Experiments for Mixtures, so you may want to investigate how they are analyzed.
Hello Oleksandra,
You can't construct a 3-value total, since (as Andrey points out), there is a linear dependence among the three values, and all totals would be 100 anyhow! You could do the following:
1. Record two of the three variables (perhaps Luck% and Effort%) for each task for each participant. The basic analysis could be one-between (groups: two levels), two-within (type of score: 2 levels--luck and effort; goal: 4 levels) variables anova.
2. Run three separate analyses (noting that there will be some redundancy since the scores are ipsative in nature): Groups (2) x Goal (4) mixed anova, with one analysis for the Luck values, a second for the effort values, and a third for the connections values.
Either approach can help you explain what differences, if any, there are as a function of group or treatment, and of goal (as well as interaction).
Good luck with your work.
Wayne Fischer
Andrey Davydenko
Now I understand more! Thank you a lot!
Your research sounds interesting.
I think you can conduct some visual analysis and use a scatterplot.
Choose different symbols for groups and different colors for goals.
Use Luck_Goal1 for X and Effort_Goal1 for Y.
Then you can show your cases using a scaterplot.
Let me recommend our framework for scatterplots:
Poster A Visual Framework for Longitudinal and Panel Studies (with ...
For a formal statistical analysis, I think the solution described above by David Morse is great.
- Badges
- Science topic