R: Observed sequential analysis.

DoOSA {aides}

R Documentation

Observed sequential analysis.

Description

DoOSA() is a function for conducting observed sequential analysis.

Usage

DoOSA(
  data = NULL,
  source = NULL,
  time = NULL,
  n = NULL,
  es = NULL,
  se = NULL,
  r1 = NULL,
  m1 = NULL,
  sd1 = NULL,
  n1 = NULL,
  r2 = NULL,
  m2 = NULL,
  sd2 = NULL,
  n2 = NULL,
  group = c("Group 1", "Group 2"),
  ref = 2,
  prefer = "small",
  measure = "ES",
  model = "random",
  method = "DL",
  pooling = "IV",
  trnsfrm = "logit",
  poolProp = "IV",
  alpha = 0.05,
  beta = 0.2,
  anchor = NULL,
  adjust = "D2",
  plot = FALSE,
  SAP = FALSE
)

Arguments

`data`	DATAFRAME consists of relevant information.
`source`	CHARACTER for labeling the included data sets.
`time`	NUMERIC values of time sequence.
`n`	INTEGER values of sample sizes.
`es`	NUMERIC values of effect sizes.
`se`	NUMERIC values of standard errors for the effect sizes.
`r1`	INTEGER values of observed events in group 1 in the included data.
`m1`	NUMERIC values of estimated means in group 1 in the included data.
`sd1`	NUMERIC values of standard deviations in group 1 in the included data.
`n1`	INTEGER values of sample sizes in group 1 in the included data.
`r2`	INTEGER values of observed events in group 2 in the included data.
`m2`	NUMERIC values of estimated means in group 2 in the included data.
`sd2`	NUMERIC values of standard deviations in group 2 in the included data.
`n2`	INTEGER values of sample sizes in group 2 in the included data.
`group`	CHARACTER for labeling two groups.
`ref`	NUMERIC values of 1 or 2 for indicating group 1 or 2 as reference.
`prefer`	CHARACTER of "small" and "large" for indicating which direction is beneficial effect in statistic test.
`measure`	CHARACTER for indicating which statistic measure should be used.
`model`	CHARACTER of "random" and "fixed" for indicating whether to use random-effects model or fixed-effect model.
`method`	CHARACTER for indicating which estimator should be used in random-effects model. In addition to the default "DL" method, the current version also supports "REML" and "PM" methods for calculating heterogeneity estimator.
`pooling`	CHARACTER for indicating which method has to be used for pooling binary data. Current version consists of "IV" and "MH" for binary data pooling.
`trnsfrm`	CHARACTER for indicating which method for transforming pooled proportion. Current version supports "none", "logit", "log", "arcsine", and "DAT" for the transformation.
`poolProp`	CHARACTER for indicating which method has to be used for pooling proportion. Current version supports "IV" and "GLMM" for the data pooling.
`alpha`	NUMERIC value between 0 to 1 for indicating the assumed type I error.
`beta`	NUMERIC value between 0 to 1 for indicating the assumed type II error.
`anchor`	NUMERIC value for indicating the presumed meaningful effect based on anchor-based approach.
`adjust`	CHARACTER for indicating how to adjust optimal information size. Current version consists of "none", "D2", "I2", "CHL", "CHM", and "CHH" for the adjustment.
`plot`	LOGIC value for indicating whether to illustrate alpha-spending monitoring plot.
`SAP`	LOGIC value for indicating whether to show sequential-adjusted power.

Details

Basic information for the function DoOSA(): DoOSA() supports observed sequential analysis of aggregate data synthesis based on head-to-head comparison using either binary or continuous data in each group. Minimum information for the function DoOSA() encompasses a data set of study-level data, and time sequence. Operative points of using function DoOSA() are listed below:

1.1. Parameter data should be used for assigning a data set.

1.2. Study-level data have to be assigned according to outcome type:

1.2.1. For dichotomous outcome: Parameter n1 and n2 should be defined with parameter r1 and r2.

1.2.2. For continuous outcome: parameter n1 and n2 should be defined with parameter m1, sd1, m2, sd2.

1.3. Parameter source and time are required for doing observed sequential analysis. Other parameters are auxiliary.

Default in the function DoOSA() Certain defaults have been elucidated in the introductory section about the parameters, but some of them need to be elaborated upon due to their complexity.

2.1. Default on the parameter measure is "ES" that automatically uses risk ratio ("RR") for binary outcome and mean difference ("MD") for continuous outcome respectively. Argument "OR" and "SMD" can be used for the parameter measure when original analysis pools data based on odds ratio or standardized mean difference.

2.2. Default on the parameter method is "DL" for applying DerSimonian-Laird heterogeneity estimator in the original pooled analysis. Other eligible arguments for the parameter are "REML" for restricted maximum-likelihood estimator, "PM" for Paule-Mandel estimator, "ML" for maximum-likelihood estimator, "HS" for Hunter-Schmidt estimator, "SJ" for Sidik-Jonkman estimator, "HE" for Hedges estimator, and "EB" for empirical Bayes estimator.

2.3. Default on the parameter pooling is "IV" for applying inverse variance weighting method. Other commonly-used and eligible arguments for the parameter are "MH" for Mantel-Haenszel method and "Peto" for pooling data using Peto method. The arguments "MH" and "Peto" are exclusively available for binary outcomes, while the argument "IV" will be automatically applied in the case of continuous outcomes.

2.4. Default on the parameter adjust is "D2" for adjusting optimal information size (OIS) based on diversity (D-squared statistics). Other eligible arguments for the parameter are "None" for the OIS without adjustment, "I2" for adjusted OIS based on I-squared statistics, "CHL" for adjusted OIS based on low heterogeneity by multiplying 1.33, "CHM" for adjusted OIS by multiplying 2 due to moderate heterogeneity, and "CHL" for adjusted OIS by multiplying 4 due to high heterogeneity.

Value

DoOSA() returns a summary on the result of sequential analysis, and can be stored as an object in DoOSA class. Explanations of returned information are listed as follows:

`studies`	Numbers of studies included in the sequential analysis.
`AIS`	Acquired information size refers to the total sample size in the sequential analysis.
`alpha`	A numeric value of type I error for the sequential analysis.
`beta`	A numeric value of type II error for the sequential analysis.
`OES`	A numeric value of observed effect size of meta-analysis.
`variance`	A numeric value of variance of meta-analysis.
`diversity`	A numeric value to show diversity in the pooled analysis.
`AF`	A numeric value of adjustment factor.
`OIS.org`	A numeric value for optimal information size without adjustment.
`OIS.adj`	A numeric value for optimal information size with adjustment.
`frctn`	A vector of fraction of each study included in the sequential analysis.
`weight`	A vector of weight of each study included in the sequential analysis.
`es.cum`	A vector of cumulative effect size in the sequential analysis.
`se.cum`	A vector of standard error for the cumulative effect size in the sequential analysis.
`zval.cum`	A vector of cumulative z-value in the sequential analysis.
`asb`	A data frame of alpha-spending values for each study.
`aslb`	A numeric value for lower alpha-spending boundary.
`asub`	A numeric value for upper alpha-spending boundary.

Author(s)

Enoch Kang

References

Jennison, C., & Turnbull, B. W. (2005). Meta-analyses and adaptive group sequential designs in the clinical development process. Journal of biopharmaceutical statistics, 15(4), 537–558. https://doi.org/10.1081/BIP-200062273.

Revicki, D., Hays, R. D., Cella, D., & Sloan, J. (2008). Recommended methods for determining responsiveness and minimally important differences for patient-reported outcomes. Journal of clinical epidemiology, 61(2), 102-109. https://doi.org/10.1016/j.jclinepi.2007.03.012.

Wetterslev, J., Jakobsen, J. C., & Gluud, C. (2017). Trial sequential analysis in systematic reviews with meta-analysis. BMC medical research methodology, 17(1), 1-18.

NCSS Statistical Software (2023). Group-sequential analysis for two proportions. In PASS Documentation. Available online: https://www.ncss.com/wp-content/themes/ncss/pdf/Procedures/NCSS/Group-Sequential_Analysis_for_Two_Proportions.pdf

Examples

## Not run:
# 1. Import a dataset of study by Fleiss (1993)
library(meta)
data("Fleiss1993bin")

# 2. Perform observed sequential analysis
output <- DoOSA(Fleiss1993bin, study, year,
                r1 = d.asp, n1 = n.asp,
                r2 = d.plac, n2 = n.plac,
                measure = "RR",
                group = c("Aspirin", "Control"))

## End(Not run)

[Package aides version 1.3.3 Index]