I am conducting a research to obtain my master diploma at the moment and I am facing a problem related to the analysis of different treatment groups.
To give you in idea about what I’m talking about: I am interested in the influence of personalization and aesthetic appearance of newsletter on mood, risk and intention to login.
Therefore, I conducted a 3x2 research (personalization: no, medium, high; and design: simple and high aesthetic design). Thus, I have 6 manipulation groups in my survey.
My aim is to analyze if there is a difference between the manipulations of personalization and between the two levels of design.
If I use the categorical variable personalization as IV for the first model, and run the pls algorithm, it cannot be calculated for the three treatment levels and only for the full data set. Same occurs if I run the MGA.
If I take the categorical IV “newsletter manipulation” (values 1 to 6), I can run the PLS algorithm for the full data and for all groups (all 3 of personalization and both of design). Thus, I could analyze the measurement model based on the aggregated data and all 5 levels of the manipulation. And I can analyze the difference between all 3x2 research design levels.
Is this a valid approach? And is it valid to include IV “number the manipulation”?
Or do I have to include a categorical predictor IV as dummy coded variable? Then I could compare, e.g., for personalization: medium compared to the reference category no personalization. And in a second model: high personalization to the reference category.
Or is none of my approaches suitable?
I hope I have explained myself clearly about my trouble and I would appreciate if anyone could give some advice.
Kind regards,
Laura Fischer