This page contains software resources to accompany the paper

- R commands for the examples in the paper
- Stata commands for graphs equivalent to those in the paper
- Simple stripcharts and violinplots can be produced online using BoxPlotR, as described by Spitzer et al
- Excel file showing how to construct Figures 1 to 5
- GraphPad PRISM can produce stacked dotcharts, multiple stacked dotcharts and scatterplots. Instructions are in the supplemental information of Weissgerber et al

In all cases, the data are random samples from the freely-available NHANES dataset, and the variables used are;

**BPXSAR**: systolic blood pressure (mmHg)**BPXDAR**: diastolic blood pressure (mmHg)**BPXDI1, BPXDI2**: two diastolic blood pressure readings**race_ethc**: race/ethnicity, coded as Hispanic, White non-Hispanic, Black non-Hispanic and Other**DR1TFOLA**: folate intake (\(\mu\)g/day)**RIAGENDR**: sex, coded as Male/Female**BMXBMI**: body mass index (kg/m\(^2\))**RIDAGEY**: age (years)

The data are available as comma-separated files; (missing values are denoted “.”)

- small dataset, n=30
- medium dataset, n=200
- large dataset, n=1000

Note: for simplicity, all statistical analyses depicted in the graphs assume that the data represent a simple random sample (i.e. independent observations) from a population of interest. Regression analyses make the standard assumptions that the mean model is correctly specified, and (for linear regression) that the variance of outcomes is constant with respect to covariates.