Hi everyone!
In my thesis, I use partner gender (woman vs. man) as a between subjects independent variable and priming figure (partner vs. acquaintance) as a within subject independent variable. Dependent variable is negative affect recovery score (range: -5 and +5).
So my study design is a 2x2 mixed design.
Also, I gave all participants LGB Identity Scale and asked their relationship duration prior to the experiment. I planned to use LGB Identity Scale (1-7 Likert type) and Relationship Duration (in months) as control variables.
The problem is that two control variables are continuous and what they measure is completely different than what I measure in the experiment? therefore, I cannot imagine how these covariates can eliminate any effect from dependent variable?
Another question is which analysis to use?
Actually I need to understand the logic of any analysis method to interpret these variables.
Thank you!
(*I use Tabachnick and Fidell’s book you can also refer to a book page )
Dear Idil Ugurluoglu ,
Your question is too complex for me to follow and it sounds like you have also managed to confuse yourself,
I would suggest adopting the principle of parsimony by starting with simple descriptive statistics and the simplest model with the maximum explanatory powerm then adding complexity to your model in stages.
These mixed factorial ANOVA designs can be very confusing and can be like a sledgehammer trying to crack a nut. You should always have a feel for your data before running a statistical analysis.
Hello Idil,
The 2 x 2 (one-between, one-within) portion of the design sounds straightforward enough. Whether the proposed covariates do or don't help depends on:
1. To what extent you have a reasonable basis for presuming that LGB identity and relationship duration would/might/could be associated with negative affect recovery (and whether this is a part of your specific research question/s);
2. To what extent the covariates do correlate with negative affect recovery; and
3. To what extent the partner gender groups differ on LBG identity and/or relationship duration.
If #2 is "zero," and #3 is "not at all," then inclusion of the covariates will not affect the analysis (except for the deduction of 2 df from the error term for between-subjects effects).
#1 is all about having a rationale for all of your variables.
Good luck with your work!
I conducted a romantic relationship attachment study with bisexual women. By this way, I wanted to test whether the affect regulatory effect of attachment differs between two groups of bisexual women (one with male partner, other with woman+ partner). Yet, collaborating with sexual minority sample made me think of Identity and I found supportive arguments in the bisexuality literature. Thus, I wanted to eliminate the noise of Identity variable whether it has any extra effect on my DV other than partner gender.
One year of relationship is claimed as a "cut-off" for building attachment in a romantic relationship. But it was too ideal for my data collection and I have a sample with diverse rel. duration (range:1-108), therefore, I thought that testing the relationship duration whether it has any interaction with my DV could be better way to understand the results of the main analysis.
Thank you for your answers! I will follow your suggestions David Morse Peter Samuels
- Badges
- Science topic