rep_consensus_group_mod: Reputation Consensus with Group Moderator Model

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

Description

This fits a model estimating the possible hearsay reputation parameters as a multi-group path model given vectors of P1 and P2 reports (vectors of quoted variable names) and a group-level categorical variable. The baseline model estimates each parameter seperately, labelling parameters based on the labels of moderator variable. Those parameters are:

hc

hearsay consensus; the correlation between P1(T) & P2(T)

int_p1

Intercept for P1(T)

int_p2

Intercept for P2(T)

v_p1

variance for P1(T)

v_p2

variance for P2(T)

p1_p2_rel_el

P1-P2 Relative Elevation (i.e., Mean P1(T) - Mean P2(T))

If n exchangeable triads > 1:

rec

direct reciprocity; the correlation between opposit P1(T)s (e.g., A(C) <-> C(A))

h

hearsay reciprocity; the correlation between exchangeable P2(T)s (e.g., B(C) <-> D(A))

m

unnamed parameter; The correlation between P2(T) and the opposite P1(T) in a group. (e.g., B(C) <-> C(A))

The function can handle up to n exchangeable triads.

Usage

rep_consensus_group_mod(data, model = NULL, p1_reports, p2_reports,
  group_mod = NULL, use_labs = TRUE, groups_eql = "none",
  params_eql = "none", n_triads = length(p1_reports),
  n_p1s_per_p2s = 1, n_p2s_per_p1s = 1)

Arguments

data	The dataframe that contains P1 & P2 ratings and the group-level moderator. Data should be wide, with a row for every group of participants. At a minimum, it must contain three columns: one for P1 reports, one for P2 reports, and one for the group-level moderator.
model	Optional. A model from the corresponding ReputationModelR model builder function. If this is supplied, no additional arguments need to be specified.
p1_reports	Quoted column names that contain P1 reports, or ratings made by the person that knows the target directly. If more than one is supplied, the target-wise order must match the other rating types.
p2_reports	Quoted column names that contain P2 reports, or ratings made by the person that knows the target indirectly through the corresponding P1. If more than one is supplied, the target-wise order must match the other rating types.
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" for hearsay consensus), or use one of several built-in options including "all" which constrains all parameters to be equal across groups and "main" which constrains just the hearsay consensus to be equal across groups (in this model).
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_p1s_per_p2s	The number of P1s for every P2. This defaults to 1. Currently, only values of 1 are supported.
n_p2s_per_p1s	The number of P2s for every P1;. This defaults to 1. Currently, only values of 1 are supported.

Value

The function returns an object of class lavaan.

Examples

data("rep_sim_data")
          agree_consensus_grpmod <- rep_consensus_group_mod(data = rep_sim_data,
                                                            p1_reports = c("A_C_agreeableness", "C_A_agreeableness"),
                                                            p2_reports = c("B_C_agreeableness", "D_A_agreeableness"),
                                                            group_mod = "study")
         # alternatively
         # build the model
          agree_consensus_grpmod_model <- rep_consensus_group_mod_builder(p1_reports = c("A_C_agreeableness", "C_A_agreeableness"),
                                                                          p2_reports = c("B_C_agreeableness", "D_A_agreeableness"),
                                                                          groups = levels(rep_sim_data$study))
         # then fit it
         agree_consensus_grpmod <- rep_consensus_group_mod(data = rep_sim_data,
                                                           model = agree_consensus_grpmod_model,
                                                           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_consensus_grpmod <- rep_consensus_group_mod(data = rep_sim_data,
                                                          p1_reports = c("A_C_agreeableness", "C_A_agreeableness"),
                                                          p2_reports = c("B_C_agreeableness", "D_A_agreeableness"),
                                                          group_mod = "study",
                                                          groups_eql = "all",
                                                          params_eql = "all")

        # Or we could constrain just hearsay consensus to be equal
         agree_consensus_grpmod <- rep_consensus_group_mod(data = rep_sim_data,
                                                       p1_reports = c("A_C_agreeableness", "C_A_agreeableness"),
                                                       p2_reports = c("B_C_agreeableness", "D_A_agreeableness"),
                                                       group_mod = "study",
                                                       groups_eql = "all",
                                                       params_eql = c("hc", "v_p1", "v_p2"))

       # It can also handle more groups, and groups without unlabelled groups.
       # The simulated dataset has the group_var variable, which is contains
       # 4 groups, each labelled with just a number (1 to 4). To use an unlabelled group
       # either set use_labs to FALSE or it will do so for you.
       agree_4grp_consensus_fit <- rep_consensus_group_mod(data = rep_sim_data,
                                                       p1_reports = c("A_C_agreeableness", "C_A_agreeableness"),
                                                       p2_reports = c("B_C_agreeableness", "D_A_agreeableness"),
                                                       group_mod = "group_var")

      # for unlabelled groups (or whenever use_labs = FALSE), you can select certain groups for equality constraints
      # by passing a vector of group numbers that should be constrained to be equal. For example, if we wanted
      # just groups 1 and 3 to be equal, we would set groups_eql to c(1, 3), like so:
       agree_4grp_consensus_fit <- rep_consensus_group_mod(data = rep_sim_data,
                                                       p1_reports = c("A_C_agreeableness", "C_A_agreeableness"),
                                                       p2_reports = c("B_C_agreeableness", "D_A_agreeableness"),
                                                       group_mod = "group_var",
                                                       groups_eql = c(1, 3),
                                                       params_eql = "all")

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