models/goodbad_models/a_c.R

In soniamitchell/SpARKjags:

# Latent variable model / hierarchical model

model {
  # Define likelihood model for data: f(y_i|theta)
  for (p in 1:N_patients) {
    # probability of belonging to the bad group (latent variable)
    bad.p[p] ~ dbern(prob.of.bad.hosp)
    # index the bad group in ac.prob
    index.bad.p[p] <- bad.p[p] + 1

    for(a in 1:antibiotic_classes) {
      # Response is different for each antibiotic and depending on
      # which pop it's from
      response[h_GUID[p],a] ~
        dbern(ac.prob[a,
                      index.bad.p[p],
                      clinical[sample_type[h_sample_GUID[p]]]])
    }
  }


  for (gp in 1:N_gp) {
    # probability of belonging to the bad group
    bad.gp[gp] ~ dbern(prob.of.bad.gp)
    # index the bad group in ac.prob
    index.bad.gp[gp] <- bad.gp[gp] + 1

    for(a in 1:antibiotic_classes) {
      # Response is different for each antibiotic and depending on
      # which pop it's from
      response[gp_GUID[gp],a] ~ dbern(a.gp.prob[a, index.bad.gp[gp]])
    }
  }

  for (v in 1:N_volunteers) {
    # probability of belonging to the bad group
    bad.v[v] ~ dbern(prob.of.bad.vol)
    # index the bad group in ac.prob
    index.bad.v[v] <- bad.v[v] + 1

    for(a in 1:antibiotic_classes)
    {
      # Response is different for each antibiotic and depending on
      # which pop it's from
      response[v_GUID[v],a] ~ dbern(a.v.prob[a, index.bad.v[v]])
    }
  }

  for (o in 1:N_outpatients) {
    # probability of belonging to the bad group
    bad.o[o] ~ dbern(prob.of.bad.out)
    # index the bad group in ac.prob
    index.bad.o[o] <- bad.o[o] + 1

    for(a in 1:antibiotic_classes)
    {
      # Response is different for each antibiotic and depending on
      # which pop it's from
      response[o_GUID[o],a] ~ dbern(a.o.prob[a, index.bad.o[o]])
    }
  }

  # ------------------------

  # Define the priors:
  for(a in 1:antibiotic_classes) {
    # probability of being resistant in the good group (less resistances)
    antibiotic.class.effect[a] ~ dnorm(intercept, tau.class)
    antibiotic.class.diff[a] ~ dnorm(diff, tau.diff)
    logit(a.prob[a,1]) <- antibiotic.class.effect[a] - antibiotic.class.diff[a] / 2

    # probability of being resistant in the bad group (many resistances) will
    # always be higher than the good group
    logit(a.prob[a,2]) <- antibiotic.class.effect[a] + antibiotic.class.diff[a] / 2

    for(b in 1:2) { # good bad
      for (c in c(ncarr, nclin)) { # 1, 2!
        ac.effect[a,b,c] ~ dnorm(antibiotic.class.effect[a,b], tau.clin)
        logit(ac.prob[a,b,c]) <- ac.effect[a,b,c]
      }

      a.gp.effect[a,b] <- ac.effect[a,b,gp_clinical]
      a.v.effect[a,b] <- ac.effect[a,b,v_clinical]
      a.o.effect[a,b] <- ac.effect[a,b,o_clinical]
      logit(a.gp.prob[a,b]) <- a.gp.effect[a,b]
      logit(a.v.prob[a,b]) <- a.v.effect[a,b]
      logit(a.o.prob[a,b]) <- a.o.effect[a,b]
    }
  }

  # Prior value for intercept
  intercept ~ dnorm(0, 0.001)

  # Difference between distribution means of "good" and "bad" groups
  diff ~ dgamma(0.001, 0.001)
  intercept.plus <- intercept + diff

  # Probability of being in the bad group
  prob.of.bad.hosp ~ dbeta(1, 1)
  prob.of.bad.gp ~ dbeta(1, 1)
  prob.of.bad.vol ~ dbeta(1, 1)
  prob.of.bad.out ~ dbeta(1, 1)

  # Prior values for precision
  tau.class ~ dgamma(0.001, 0.001)
  tau.clin ~ dgamma(0.001, 0.001)
  tau.diff ~ dgamma(0.001, 0.001)

  # Convert precisions to sd
  sd.class <- sqrt(1/tau.class)
  sd.clin <- sqrt(1/tau.clin)
  sd.diff <- sqrt(1/tau.diff)

  #monitor# full.pd, dic, deviance, a.prob, ac.prob, a.gp.prob, a.v.prob, a.o.prob, prob.of.bad.hosp, prob.of.bad.gp, prob.of.bad.vol, prob.of.bad.out, bad.p, bad.gp, bad.v, bad.o, intercept, sd.class, sd.clin, sd.diff
}