Fit gastric emptying curves with Stan

stan_gastempt(d, model_name = "linexp_gastro_2b", lkj = 2,
  student_df = 5L, init_r = 0.2, chains = 4, ...)

Arguments

d

A data frame with columns

  • rec Record descriptor as grouping variable, e.g. patient ID

  • minute Time after meal or start of recording.

  • vol Volume of meal or stomach

model_name

Name of predefined model in gastempt/exec. Use stan_model_names() to get a list of available models.

lkj

LKJ prior for kappa/tempt correlation, only required for model linexp_gastro_2b. Values from 1.5 (strong correlation) to 50 (almost independent) are useful. See http://www.psychstatistics.com/2014/12/27/d-lkj-priors/ for examples.

student_df

Student-t degrees of freedom for residual error; default 5. Use 3 for strong outliers; values above 10 are close to gaussian residual distribution.

init_r

for stan, default = 0.2; Stan's own default is 2, which often results in stuck chains.

chains

for stan; default = 4. For debugging, use 1.

...

Additional parameter passed to sampling

Value

A list of class stan_gastempt with elements coef, fit, plot

  • coef is a data frame with columns:

    • rec Record descriptor, e.g. patient ID

    • v0 Initial volume at t=0

    • tempt Emptying time constant

    • kappa Parameter kappa for model = linexp

    • beta Parameter beta for model = powexp

    • t50 Half-time of emptying

    • slope_t50 Slope in t50; typically in units of ml/minute On error, coef is NULL

  • fit Result of class `stanfit`

  • plot A ggplot graph of data and prediction. Plot of raw data is returned even when convergence was not achieved.

Examples

# NOT RUN {
  dd = simulate_gastempt(n_records = 6, seed = 471)
  d = dd$data
  ret = stan_gastempt(d)
  print(ret$coef)
# }