make_data: Create simulated datasets.

In dhaine/misclass: Sampling Strategies to Control Misclassification Bias in Longitudinal Udder Health Studies

Description

Create datasets from a hypothetical cohort study. Since correlation structures are often found in udder health studies, the generated datasets are deemed to have occurred from the collection of two milk samples collected 1 month apart from each quarters of a random sample of 30 cows per herd, from 100 dairy herds. The first milk sample (S1) is used to identify quarters at risk of intramammary infection (IMI) at the beginning of the cohort, while the second (S2) is used to identify the outcome (acquisition of a new IMI). Three hypothetical exposures E_q, E_c, E_h (quarter, cow, and herd level) with known strength of association (OR~3.0) are generated. As it is often the case (Dufour et al., 2012), exposures are equally associated with odds of a prevalent IMI on first milk sample as with odds of IMI acquisition on second sample. Exposures are randomly associated with odds of eliminating an existing IMI (OR=1.0). If S. aureus or CNS is chosen, default parameters are used. Otherwise user has to provide his own.

Usage

make_data(n_herd, n_cow, bact = c("saureus", "cns", "other"), E_hPr = NULL,
  E_cPr = NULL, E_qPr = NULL, sigma_sqhPr = NULL, sigma_sqcPr = NULL,
  b0_Pr = NULL, OR_hPr = NULL, OR_cPr = NULL, OR_qPr = NULL,
  sigma_sqhI = NULL, sigma_sqcI = NULL, b0_I = NULL, OR_hI = NULL,
  OR_cI = NULL, OR_qI = NULL, sigma_sqhEl = NULL, sigma_sqcEl = NULL,
  b0_El = NULL, se_parms = NULL, sp_parms = NULL, se_series = NULL,
  sp_series = NULL, se_parall = NULL, sp_parall = NULL, se_tri = NULL,
  sp_tri = NULL)

Arguments

n_herd	Number of herds.
n_cow	Number of cows per herd.
bact	Type of bacteria: S. aureus, CNS, or other. If other, as to provide parameters.
E_hPr	Exposure distribution (0 to 1) of the binary herd-level (h) predictor. S. aureus and CNS = 0.5.
E_cPr	Exposure distribution (0 to 1) of the binary cow-level (c) predictor. S. aureus and CNS = 0.5.
E_qPr	Exposure distribution (0 to 1) of the binary quarter-level (q) predictor. S. aureus and CNS = 0.5.
sigma_sqhPr	Herd-level variance (sigma_sq) for prevalence of intra-mammary infection (IMI). S. aureus = 0.14; CNS = 0.363.
sigma_sqcPr	Cow-level variance for prevalence of IMI. S. aureus = 2.25; CNS = 0.294.
b0_Pr	Intercept for IMI prevalence; aiming at a prevalence of 2.5%. S. aureus = -6.7; CNS = -2.15.
OR_hPr	OR of association between herd-level variable and IMI prevalence. S. aureus and CNS = 3.
OR_cPr	OR of association between cow-level variable and IMI prevalence. S. aureus and CNS = 3.
OR_qPr	OR of association between observation-level variable and IMI prevalence. S. aureus and CNS = 3.
sigma_sqhI	Herd-level variance for incidence of IMI. S. aureus = 0.838; CNS = 0.27.
sigma_sqcI	Cow-level variance for incidence of IMI. S. aureus = 2.926; CNS = 0.256.
b0_I	Intercept for IMI incidence. S. aureus = -8.3; CNS = -2.4.
OR_hI	OR of association between herd-level variable and IMI incidence. S. aureus and CNS = 3.
OR_cI	OR of association between cow-level variable and IMI incidence. S. aureus and CNS = 3.
OR_qI	OR of association between observation-level variable and IMI incidence. S. aureus and CNS = 3.
sigma_sqhEl	Herd-level variance for elimination of IMI. S. aureus = 0.15; CNS = 0.112.
sigma_sqcEl	Cow-level variance for elimination of IMI. S. aureus = 2.246; CNS = 0.7.
b0_El	Intercept for IMI persistency. S. aureus = -0.6; CNS = 1.6.
se_parms	Vector for mode and x to determine shape parameters of Beta distribution of sensitivity (Se). S. aureus = c(0.90, 0.85); CNS = c(0.60, 0.55).
sp_parms	Vector for mode and x to determine shape parameters of Beta distribution of specificity (Sp). S. aureus = 1; CNS = c(0.95, 0.90).
se_series	Se improvement or loss associated with sampling strategy: duplicate series. S. aureus = -0.1; CNS = -0.25.
sp_series	Sp improvement or loss associated with sampling strategy: duplicate series. S. aureus = 0; CNS = 0.05.
se_parall	Se improvement or loss associated with sampling strategy: duplicate parallel. S. aureus = 0.1; CNS = 0.15.
sp_parall	Sp improvement or loss associated with sampling strategy:duplicate parallel. S. aureus = 0; CNS = -0.05.
se_tri	Se improvement or loss associated with sampling strategy: triplicate (2 out of 3). S. aureus = 0; CNS = 0.
sp_tri	Sp improvement or loss associated with sampling strategy: triplicate (2 out of 3). S. aureus = 0; CNS = 0.10.

Value

A data frame with variables:

herd	Herd id.
cow	Cow id.
quarter	Quarter id.
S1	First milk sample true status.
S2	Second milk sample true status.
E_h	Herd-level exposure.
E_c	Cow-level exposure.
E_q	Quarter-level exposure.
S1i	Misclassified first milk sample.
S2i	Misclassified second milk sample.
S1_series	Misclassified first milk sample based on duplicate series sampling strategy.
S2_series	Misclassified second milk sample based on duplicate series sampling strategy.
S1_parall	Misclassified first milk sample based on duplicate parallel sampling strategy.
S2_parall	Misclassified second milk sample based on duplicate parallel sampling strategy.
S1_tri	Misclassified first milk sample based on triplicate sampling strategy.
S2_tri	Misclassified second milk sample based on triplicate sampling strategy.

References

Dufour, S., Dohoo, I.R., Barkema, H.W., DesCôteaux, L., DeVries, T.J., Reyher, K.K., Roy, J.-P., Scholl, D.T., 2012 Epidemiology of coagulase-negative staphylococci intramammary infection in dairy cattle and the effect of bacteriological culture misclassification. Journal of Dairy Science 95(6):3110-3124.

Examples

# Initiate a list to store the n data frames
sim_list <- vector("list", 5)
# Do not forget to set seed for replication
set.seed(123)
sim_list <- replicate(n = 5, expr = make_data(100, 30, "saureus"), simplify = FALSE)
# Or with a progress bar
require(pbapply)
sim_list <- pbreplicate(n = 5, expr = make_data(100, 30, "cns"), simplify = FALSE)

dhaine/misclass documentation built on May 15, 2019, 8:21 a.m.