rep_generic_group_id_mods: Generic Individual-Level Moderator and Group Moderators

In Coryc3133/ReputationAnalyses: Functions to model and analyze triadic (T-P1-P2) reputational data

Description

This is a generic function for fitting a lavaan model for individual-level moderators on two distinguishable ratings on the same target across multiple groups. This could be P1- and P2- reports, P2- and self-reports, P1- and self-reports, or any other sets of distinguishable ratings.

Usage

rep_generic_group_id_mods(data, model = NULL, rating_1, rating_2,
  id_mod_variable, interaction_term, group_mod = NULL, use_labs = TRUE,
  groups_eql = "none", params_eql = "none",
  n_triads = length(rating_1), n_r1_per_r2 = 1, n_r2_per_r1 = 1)

Arguments

data	The dataframe. Data should be wide, with a row for every group of participants. At a minimum, it must contain five columns: one for rating 1, one for rating 2, one for the individual difference moderator, one for the interaction term, and one for the group moderated variable.
model	Optional. A model from the corresponding ReputationModelR model builder function. If this is supplied, no additional arguments need to be specified.
rating_1	Quoted column names that contain the first rating variable. This might be P1 reports if investigating moderation of hearsay consensus or self-reports for moderation of hearsay accuracy. If more than one is supplied, the target-wise order must match across variables.
rating_2	Quoted column names that contain second rating variable. For hearsay consensus or accuracy, this would be P2 reports. If more than one is supplied, the target-wise order must match across variables.
id_mod_variable	Quoted column names that contain the individual-level moderator of interest. If more than one is supplied from multiple exchangeable dyads/triads, the order must match the order of the ratings. Like P2-reports, the variable should be mean-centered to facilitate interpretability.
interaction_term	Quoted column names that contain the interaction term, or the product of the mean-centered P2-report and the mean-centered moderator variable. If more than one is supplied from multiple exchangeable dyads/triads, the target-wise order must match the order of the ratings.
group_mod	The quoted column name that contains a group-level categorical moderator.
use_labs	Logical indicating whether or not to use the group labels to create the parameter labels. If FALSE, generic labels (grp1 to grpk, where k is the number of groups) are used.
groups_eql	Optional. Groups that you want to constrain to be equal across some or all parameters. If you have use_labs set to TRUE, provide a vector of group labels corresponding to the groups you want to constrain to be equal. If you have use_labs set to FALSE, provide a vector of numbers corresponding to the position of the groups you want to constrain to be equal. If you provide "all", all groups will be constrained to be equal.
params_eql	Optional. Parameters that you want to constrain to be equal across the groups specified in groups_eql. You can provide one or more specific parameters (e.g., "hc_me" for hearsay consensus main effect), or use a built-in options including "all" which constrains all parameters to be equal across groups.
n_triads	The number of exchangeable triads in each group. By default, this is determined by counting the number of P1 reports. This parameter rarely needs to be changed.
n_r1_per_r2	The number of first ratings for each second rating. Currently, only 1:1 is supported.
n_r2_per_r1	The number of second ratings for each first rating. Currently, only 1:1 is supported.

Details

The parameters estimated by this model are:

rating_me

main effect of other rating; this should correspond to correlation between ratings at average level of moderator variable (if data were properly mean-centered).

mod_me

The meain effect of the moderator variable; it can be interpreted as the difference in rating_1 to differences in the individual-level moderator variable.

interaction

This is the interaction term. It indicates the extent to which the correlation between ratings depends on the moderator variable

v_rating_1

variance for first rating

v_rating_2

variance for second rating

v_mod

variance for moderator variable

v_interaction

variance for interaction term

int_rating_1

intercept for first rating

int_rating_2

intercept for second rating

int_mod

intercept for moderator variable

int_interaction

intercept for interaction term

The function can handle up to n exchangeable triads.

Value

The function returns an object of class lavaan.

Examples

data("rep_sim_data")
          # Prepare data
         moderator_data <- rep_sim_data %>%
           dplyr::mutate(B_C_agreeableness_cent = scale(B_C_agreeableness, scale = FALSE),
                  D_A_agreeableness_cent = scale(D_A_agreeableness, scale = FALSE),
                  B_iri_perspective_cent = scale(B_iri_perspective, scale = FALSE),
                  D_iri_perspective_cent = scale(D_iri_perspective, scale = FALSE),
                  B_ptXagree_interaction = B_C_agreeableness_cent*B_iri_perspective_cent,
                  D_ptXagree_interaction = D_A_agreeableness_cent*D_iri_perspective_cent)

# build a bsaeline model examining perspective taking moderating hearsay consensus across groups (studies in this case)
agree_pt_mod_model <- rep_generic_group_id_mods (moderator_data,
                                           rating_1 = c("C_C_agreeableness", "A_A_agreeableness"),
                                           rating_2 = c("B_C_agreeableness_cent", "D_A_agreeableness_cent"),
                                           id_mod_variable = c("B_iri_perspective_cent", "D_iri_perspective_cent"),
                                           interaction_term = c("B_ptXagree_interaction", "D_ptXagree_interaction"),
                                           group_mod = "study")

         # fit model with group equality constraints:
         # if we wanted to constrain all parameters to be equal across the 2 groups, that can be done
         # by setting groups_eql and params_eql both to "all".
agree_pt_mod_model <- rep_generic_group_id_mods (moderator_data,
                                           rating_1 = c("C_C_agreeableness", "A_A_agreeableness"),
                                           rating_2 = c("B_C_agreeableness_cent", "D_A_agreeableness_cent"),
                                           id_mod_variable = c("B_iri_perspective_cent", "D_iri_perspective_cent"),
                                           interaction_term = c("B_ptXagree_interaction", "D_ptXagree_interaction"),
                                           group_mod = "study", groups_eql = "all", params_eql = "all")

Coryc3133/ReputationAnalyses documentation built on July 31, 2019, 8:35 a.m.