simulations/irr_dv_simulation_base.R

   1 library(matrixStats) # for numerically stable logsumexps
   2
   3 options(amelia.parallel="no",
   4         amelia.ncpus=1)
   5 library(Amelia)
   6
   7 source("measerr_methods.R") ## for my more generic function.
   8
   9 run_simulation_depvar <- function(df, result, outcome_formula = y ~ x + z, rater_formula = y.obs ~ x, proxy_formula = w_pred ~ y){
  10
  11     accuracy <- df[,mean(w_pred==y)]
  12     result <- append(result, list(accuracy=accuracy))
  13
  14     (model.true <- glm(y ~ x + z, data=df, family=binomial(link='logit')))
  15     true.ci.Bxy <- confint(model.true)['x',]
  16     true.ci.Bzy <- confint(model.true)['z',]
  17
  18     result <- append(result, list(Bxy.est.true=coef(model.true)['x'],
  19                                   Bzy.est.true=coef(model.true)['z'],
  20                                   Bxy.ci.upper.true = true.ci.Bxy[2],
  21                                   Bxy.ci.lower.true = true.ci.Bxy[1],
  22                                   Bzy.ci.upper.true = true.ci.Bzy[2],
  23                                   Bzy.ci.lower.true = true.ci.Bzy[1]))
  24
  25
  26
  27     loa0.feasible <- glm(y.obs.0 ~ x + z, data = df[!(is.na(y.obs.0))], family=binomial(link='logit'))
  28
  29     loa0.ci.Bxy <- confint(loa0.feasible)['x',]
  30     loa0.ci.Bzy <- confint(loa0.feasible)['z',]
  31
  32     result <- append(result, list(Bxy.est.loa0.feasible=coef(loa0.feasible)['x'],
  33                                   Bzy.est.loa0.feasible=coef(loa0.feasible)['z'],
  34                                   Bxy.ci.upper.loa0.feasible = loa0.ci.Bxy[2],
  35                                   Bxy.ci.lower.loa0.feasible = loa0.ci.Bxy[1],
  36                                   Bzy.ci.upper.loa0.feasible = loa0.ci.Bzy[2],
  37                                   Bzy.ci.lower.loa0.feasible = loa0.ci.Bzy[1]))
  38
  39
  40     df.loa0.mle <- copy(df)
  41     df.loa0.mle[,y:=y.obs.0]
  42     loa0.mle <- measerr_mle_dv(df.loa0.mle, outcome_formula=outcome_formula, proxy_formula=proxy_formula)
  43     fisher.info <- solve(loa0.mle$hessian)
  44     coef <- loa0.mle$par
  45     ci.upper <- coef + sqrt(diag(fisher.info)) * 1.96
  46     ci.lower <- coef - sqrt(diag(fisher.info)) * 1.96
  47
  48     result <- append(result, list(Bxy.est.loa0.mle=coef['x'],
  49                                   Bzy.est.loa0.mle=coef['z'],
  50                                   Bxy.ci.upper.loa0.mle = ci.upper['x'],
  51                                   Bxy.ci.lower.loa0.mle = ci.lower['x'],
  52                                   Bzy.ci.upper.loa0.mle = ci.upper['z'],
  53                                   Bzy.ci.lower.loa0.mle = ci.upper['z']))
  54
  55     loco.feasible <- glm(y.obs.0 ~ x + z, data = df[(!is.na(y.obs.0)) & (y.obs.1 == y.obs.0)], family=binomial(link='logit'))
  56
  57     loco.feasible.ci.Bxy <- confint(loco.feasible)['x',]
  58     loco.feasible.ci.Bzy <- confint(loco.feasible)['z',]
  59
  60     result <- append(result, list(Bxy.est.loco.feasible=coef(loco.feasible)['x'],
  61                                   Bzy.est.loco.feasible=coef(loco.feasible)['z'],
  62                                   Bxy.ci.upper.loco.feasible = loco.feasible.ci.Bxy[2],
  63                                   Bxy.ci.lower.loco.feasible = loco.feasible.ci.Bxy[1],
  64                                   Bzy.ci.upper.loco.feasible = loco.feasible.ci.Bzy[2],
  65                                   Bzy.ci.lower.loco.feasible = loco.feasible.ci.Bzy[1]))
  66
  67
  68     df.loco.mle <- copy(df)
  69     df.loco.mle[,y.obs:=NA]
  70     df.loco.mle[(y.obs.0)==(y.obs.1),y.obs:=y.obs.0]
  71     df.loco.mle[,y.true:=y]
  72     df.loco.mle[,y:=y.obs]
  73     print(df.loco.mle[!is.na(y.obs.1),mean(y.true==y,na.rm=TRUE)])
  74     loco.mle <- measerr_mle_dv(df.loco.mle, outcome_formula=outcome_formula, proxy_formula=proxy_formula)
  75     fisher.info <- solve(loco.mle$hessian)
  76     coef <- loco.mle$par
  77     ci.upper <- coef + sqrt(diag(fisher.info)) * 1.96
  78     ci.lower <- coef - sqrt(diag(fisher.info)) * 1.96
  79
  80     result <- append(result, list(Bxy.est.loco.mle=coef['x'],
  81                                   Bzy.est.loco.mle=coef['z'],
  82                                   Bxy.ci.upper.loco.mle = ci.upper['x'],
  83                                   Bxy.ci.lower.loco.mle = ci.lower['x'],
  84                                   Bzy.ci.upper.loco.mle = ci.upper['z'],
  85                                   Bzy.ci.lower.loco.mle = ci.upper['z']))
  86
  87     print(rater_formula)
  88     print(proxy_formula)
  89
  90     ## mle.irr <- measerr_irr_mle( df, outcome_formula = outcome_formula, rater_formula = rater_formula, proxy_formula=proxy_formula, truth_formula=truth_formula)
  91
  92     ## fisher.info <- solve(mle.irr$hessian)
  93     ## coef <- mle.irr$par
  94     ## ci.upper <- coef + sqrt(diag(fisher.info)) * 1.96
  95     ## ci.lower <- coef - sqrt(diag(fisher.info)) * 1.96
  96
  97     ## result <- append(result,
  98     ##                  list(Bxy.est.mle = coef['x'],
  99     ##                       Bxy.ci.upper.mle = ci.upper['x'],
 100     ##                       Bxy.ci.lower.mle = ci.lower['x'],
 101     ##                       Bzy.est.mle = coef['z'],
 102     ##                       Bzy.ci.upper.mle = ci.upper['z'],
 103     ##                       Bzy.ci.lower.mle = ci.lower['z']))
 104
 105     return(result)
 106
 107 }