Cross-screening in observational studies that test many hypotheses

Author Contributions Checklist Form

Data

Abstract (Mandatory)

We use a dataset created from the National Health and Nutrition Examination Survey (NHANES) 2013-2014.

Availability (Mandatory)

The dataset is publicly available, and the subset we use is included in the R package CrossScreening available on CRAN.

Description (Mandatory if data available)

NHANES is a publicly available dataset.

• Link to data: https://wwwn.cdc.gov/nchs/nhanes/continuousnhanes/default.aspx?BeginYear=
• File format: The dataset we use can be directly read in R after the package is installed.

Code

Abstract (Mandatory)

We implement the Cross-screening method described in the submitted manuscript in R.

Description (Mandatory)

• How delivered: R package publicly available on CRAN.
• Licensing information: GPL-
• Version information: R package version 0.1.

Instructions for Use

Reproducibility (Mandatory)

Table 1 and Table 5 can be reproduced from the CrossScreening package. After installing the package and loading it to R, for Table 1 run

example(cross.screen.fg)

Note that this only reproduces the Γ=9 block in the table. For other blocks, just change 9 in the example to other values.

Table 5 can be reproduced by calling an internal function

CrossScreening:::table5(no.sims = 1)

Since Table 5 is a simulation result with 10000 simulations, it will take a long time and a lot of CPU resources to reproduce the entire table.

Replication (Optional)

The R package can be easily used for other applications. Instructions and examples are provided in the package manual and vignette.