gen_y_mmrem: gen_y_mmrem

In tessaleejohnson/corclus: Simulate Data from the Correlated Cluster Multiple Membership Model

gen_y_mmrem

Description

This function calculates the observed y value from a multiple membership random effects model based on the value of a predictor, x (given in .design_x), population values of the model coefficients (given in .gamma_x), school residuals for the schools assigned to person i (given in .sch_resid), weights for those school residuals (given in .sch_weight), and person residuals. This population model has fixed slopes and random intercepts. School-level covariates are not included in this model but are used to construct the school residuals in gen_u_mmrem.

Usage

gen_y_mmrem(.dat, .x_predictor, .sch_weight, .sch_resid, .per_resid, .gamma_x)

Arguments

.x_predictor	Numeric matrix with dimensions n x p (where n is the num of persons and p is the number of coefficients, including the intercept). The column of .design_x corresponding to the intercept should be a column of 1s.
.sch_weight	Numeric matrix with dimensions n x h (where n is the number of persons and h is the maximum number of schools attended by any person in the dataset). Rows should sum to 1 (that is, for each student, the weights assigned to their schools attended should sum to 1). For a school a student did not attend, the weight should be 0 (that is, if the maximum number of schools attended was 2 and person A only attended 1 school, then the weight for their "second school" should be 0, while the weight for their "first school" should be 1). To simulate the data, all students were initially assigned a mobility profile (meaning that all students were assigned h schools to attend), and then only a certain proportion of students were coded as mobile. For the students who were coded as mobile, their .sch_weight matrix row should give equal weight to all schools attended. For students who were coded as non-mobile, their first school was given a weight of 1 and all other schools were given weights of 0.
.sch_resid	Numeric matrix with dimensions n x h (where n is the number of persons and h is the maximum number of schools attended by any person in the dataset). The hth column of the matrix should give the residual for the hth school attended by person i. As mentioned above, all students were initially assigned a mobility profile that included multiple schools, then only a certain proportion of those mobility profiles were retained.
.per_resid	Numeric vector with length n (where n is the number of persons in the data). Gives the person-level residual for the model.
.gamma_x	Numeric vector with length p (where p is the number of model coefficients, including the intercept).

Details

NOTE: function has been changed and documentation requires an update

Value

This function returns a vector of length n (where n is the number of persons in the dataset) of y values calculated based on the population MMREM model. This population model does not explicitly model correlation between school residuals, nor does it account for school-level predictors. Correlations between school residuals and school-level predictors, if desired, must be incorporated into the .sch_resid values.

Examples

## Not run: 

## create sample values

# gamma vector (intercept, g00 & effect of x on y, g10)
g <- c(10, 1.5)

# x & design matrix
x <- rnorm(5)
d_x <- cbind(
  xg00 = rep(1, length(x)),
  xg10 = x
)

# school weight matrix
s_wt <- cbind(
  w1 = c(1, 1, 0.5, 0.5, 0.5),
  w2 = c(0, 0, 0.5, 0.5, 0.5)
)

# school residuals matrix (uncorrelated residuals)
s_r <- cbind(
  s1 = rnorm(5),
  s2 = rnorm(5)
)

# person residual vector
p_r <- rnorm(5)

## generate y values from the population model
gen_y_mmrem(
  .gamma_x = g,
  .x_predictor = d_x,
  .sch_weight = s_wt,
  .sch_resid = s_r,
  .per_resid = p_r
)


## End(Not run)

tessaleejohnson/corclus documentation built on Oct. 11, 2022, 3:46 a.m.