Analytic Plan
-------------
H1: On the basis of previous findings, and preliminary results, we expect that partners' solitary pornography use will moderate the association between actors' solitary pornography use and relationship satisfaction such that satisfaction will be lower when actors and partners are discrepant in their pornography use. Further, we expect this effect to emerge even after controlling for shared pornography use.
H2: On the basis of previous findings, and preliminary results, we expect that shared pornography use (aggregated across partner estimates) will be positively correlated with relationship satisfaction after controlling for both partners' solitary pornography use as well as the interaction between such use.
- Analysis: H1 & H2 will be tested simultaneously with a multilevel modeling (MLM) approach to the
actor-partner model (e.g., Kenny, Kashy, & Cook,2006), our initial
model (Model 1) will involve the prediction of relationship satisfaction
using an intercept, actors' and partners' attitudes towards
pornography, and their interaction, as well as a couples shared pornography use (aggregated across couples) as fixed effects. This model will
include a consideration of couple by actor and couple by partner
random effects, but these elements will be dropped if they do not
improve fit. The residual matrix will nest participant crossed with
time within couples using a heterogenous first-order autoregressive
structure to allow the residuals to correlate across measurement
occasions and partners.
H3: Despite the indications of gendered effects in the preliminary analyses outlined above, previous findings across three datasets consistently indicate that the relationships in H1 and H2 are not moderated by gender. Given the relatively small gender differences described in the preliminary analyses, we suspect that formal tests for moderation of the associations in H1 and H2 by gender will result in null effects.
- Analysis: A consideration of gender will begin with the Model 1
outlined above and will add gender, as well as interactions between gender and all of the other fixed effects (Model 2). H3 will be
tested by comparing the change in model fit between Model 1 and Model 2. If there are significant gender effects, random couple effects will also be considered for significant gendered components. The
residual matrix will remain the same as in Model 1.
H4: We previously found that the interaction between actors' and partners' attitudes towards pornography at baseline was not significantly moderated by time when predicting relationship satisfaction over time. If the effects of concordant vs. discordant pornography use are driven by similarities / dissimilarities in attitudes, we should expect a similar temporal pattern in the current sample. That is, the expected association between relationship satisfaction (measured at each wave) and the interaction between actors' and partners' pornography use (measured at T1) should not be moderated by time. We believe that the temporal stability in the interaction effect likely reflects relative consistency of pornography-related attitudes and behaviors over time, in which case, the average contextual association between relationship satisfaction and the interaction between actors' and partners' pornography use at each wave should also not be moderated by time.
- Analysis: H4 will be tested with two models. Model 3 will begin with
either Model 1 or Model 2, depending on the outcome of H3.
Regardless, the measurement level fixed effects will be removed and
replaced with partner level fixed effects representing the couples'
patterns of pornography use on the first measurement occasion (T1).
Time components will be added to this model, including a main effect
for time, and interactions between time and the other components in
the model. If the model fit improves with the addition of the time
components, then relevant random couple effects will also be
considered. The residual matrix will remain the same as in Model 1.
This model is more comparable to the temporal model that we used when
examining attitudes towards pornography (see [here][1]). To further test
H4, Model 4 will also begin with either Model 1 or Model 2, depending
on the outcome of H3. To this model we will add time components as
fixed effects, including a main effect for time, and
interactions between time and the other components in the model. If
the model fit improves with the addition of the time components, then
relevant random couple effects will also be considered. The residual
matrix will remain the same as in Model 1. This model is included to
determine if time moderates the context specific associations between
pornography use and relationship satisfaction at each measurement
occasion.
H5-H8: The pattern of relationships described above are also expected to emerge when predicting sexual satisfaction. That is, sexual satisfaction should be associated with an interaction between actors' and partners' pornography use (H5), and couples shared use of pornography (H6). Further, these associations should not be moderated by gender (H7) or time (H8). The nature of the relationships outlined in H5 and H6 will vary in direction depending on which operationalization of sexual satisfaction is used (i.e. sexual satisfaction vs. sexual dissatisfaction).
- Analyses: H5-H8 will be tested in the same way that we test H1-H4 (outlined above) with measures of sexual satisfaction serving as the dependent variables rather than relationship satisfaction. Models 1-4 will be run once for each of 3 sexual satisfaction measures: sexual satisfaction items; sexual dissatisfaction items; and a composite of sexual satisfaction and dissatisfaction items.
R1: In our previous samples, we have been unable to find quadratic effects along the line of incongruence (see Shannock et al. 2010) between actors' and partners' pornography use, and with respect to attitudes towards pornography, we found an unexpected quadratic effect along the line of congruence. Given the inconsistencies in these results, it is unclear whether quadratic effects along the lines of congruence / incongruence will emerge in the current sample, but they will be investigated to further understand the phenomenon under study.
- Analyses: To explore this issue, we will conduct a Response Surface
Analysis by adding quadratic actor and partner effects of solitary
pornography use to Model 1 (separately for relationship and sexual
satisfaction). The curvilinear effects along the lines of congruence and incongruence,
and related significance tests, will be calculated as per the
formulas outlined in Shannock et al. (2010).
[1]: https://osf.io/3wxyd/