Type:	Package
Title:	Estimate Sample Size Based on Precision Rather than Power
Version:	0.1.2
Maintainer:	Malcolm Barrett <malcolmbarrett@gmail.com>
Description:	Estimate sample size based on precision rather than power. 'precisely' is a study planning tool to calculate sample size based on precision. Power calculations are focused on whether or not an estimate will be statistically significant; calculations of precision are based on the same principles as power calculation but turn the focus to the width of the confidence interval. 'precisely' is based on the work of 'Rothman and Greenland' (2018).
License:	MIT + file LICENSE
URL:	https://github.com/malcolmbarrett/precisely
BugReports:	https://github.com/malcolmbarrett/precisely/issues
Depends:	R (≥ 3.2.0)
Imports:	dplyr, ggplot2, magrittr, purrr, rlang, shiny, shinycssloaders, shinythemes, tidyr
Suggests:	covr, ggrepel, knitr, rmarkdown, spelling, testthat, vdiffr
VignetteBuilder:	knitr
Encoding:	UTF-8
Language:	en-US
RoxygenNote:	7.1.2
NeedsCompilation:	no
Packaged:	2021-10-10 01:18:50 UTC; malcolmbarrett
Author:	Malcolm Barrett [aut, cre]
Repository:	CRAN
Date/Publication:	2021-10-10 04:30:02 UTC

Pipe operator

Description

See magrittr::%>% for details.

Usage

lhs %>% rhs

Launch precisely Shiny app

Description

launch_precisely_app() launches a Shiny app to calculate and plot precision, sample size, and upper limit calculations.

Usage

launch_precisely_app()

Calculate with precisely functions across values

Description

map_precisely() is a wrapper around tidyr::crossing() and purrr::pmap_dfr() to give a set of values to any of the calculation functions in precisely. All possible combinations of the values are passed to the function, returning a tibble where each row is the result for each combination.

Usage

map_precisely(.f, ...)

Arguments

Value

a tibble

Examples

map_precisely(
  n_risk_difference,
  precision = seq(from = .02, to = .20, by = .005),
  exposed = c(.2, .4),
  unexposed = c(.1, .3),
  group_ratio = 1
)

Estimate sample size based on precision of a measure

Description

These functions calculate the sample size needed to estimate a measure with a certain precision. For ratio measures, like the risk ratio, rate ratio, and odds ratio, this is the ratio of the upper to lower limit of the confidence interval. For difference measures, like the risk difference or rate difference, this is the absolute width of the confidence interval.

Usage

n_risk_difference(precision, exposed, unexposed, group_ratio, ci = 0.95)

n_risk_ratio(precision, exposed, unexposed, group_ratio, ci = 0.95)

n_rate_difference(precision, exposed, unexposed, group_ratio, ci = 0.95)

n_rate_ratio(precision, exposed, unexposed, group_ratio, ci = 0.95)

n_odds_ratio(
  precision,
  exposed_cases,
  exposed_controls,
  group_ratio,
  ci = 0.95
)

Arguments

Value

a tibble with sample size, effect measure, and precision

References

Rothman, K.J. and Greenland, S. 2018. Planning Study Size Based on Precision Rather Than Power. 29(5):599-603.

Examples

# From Rothman and Greenland 2018

n_risk_difference(
  precision = .08,
  exposed = .4,
  unexposed = .3,
  group_ratio = 3,
  ci = .90
)

n_risk_ratio(
  precision = 2,
  exposed = .4,
  unexposed = .3,
  group_ratio = 3
)

Plot precisely

Description

Simple line plots for the output of map_precisely(). Use dplyr::group_by() to create multiple lines on the plot.

Usage

plot_sample_size(.df, xlab = "Sample Size", ylab = "Precision", line_size = 1)

plot_precision(.df, xlab = "Precision", ylab = "Sample Size", line_size = 1)

plot_upper_limit(
  .df,
  xlab = "Sample Size",
  ylab = "Upper Limit",
  line_size = 1
)

Arguments

Value

a ggplot

Examples

library(dplyr)
library(ggplot2)

map_precisely(
  n_risk_difference,
  precision = seq(from = .02, to = .20, by = .005),
  exposed = .4,
  unexposed = .3,
  group_ratio = 1
 ) %>%
  plot_sample_size()

map_precisely(
  precision_odds_ratio,
  n_cases = seq(from = 500, to = 1000, by = 10),
  exposed_cases = .6,
  exposed_controls = .4,
  group_ratio = 1:4
) %>%
  group_by("Control/Case Ratio" = factor(group_ratio)) %>%
  plot_precision()

map_precisely(
  upper_rate_ratio,
  upper_limit = seq(1.5, 2.5, by = .1),
  prob = seq(.50, .95, by = .05),
  exposed = .01,
  unexposed = .01,
  group_ratio = 1:4
 ) %>%
  group_by("Probability" = factor(prob)) %>%
  plot_upper_limit(line_size = 1) +
    scale_color_viridis_d() +
    theme_precisely() +
    theme(legend.position = "right",
          strip.text = element_text(margin = margin(b = 5), hjust = 0)) +
    facet_wrap(~ group_ratio,
               labeller = as_labeller(function(x) paste("Unexposed/Exposed:", x)))

Estimate precision of a measure based on sample size

Description

These functions calculate the precision of an estimate given a certain sample size. For ratio measures, like the risk ratio, rate ratio, and odds ratio, this is the ratio of the upper to lower limit of the confidence interval. For difference measures, like the risk difference or rate difference, this is the absolute width of the confidence interval.

Usage

precision_risk_difference(
  n_exposed,
  exposed,
  unexposed,
  group_ratio,
  ci = 0.95
)

precision_risk_ratio(n_exposed, exposed, unexposed, group_ratio, ci = 0.95)

precision_rate_difference(
  n_exposed,
  exposed,
  unexposed,
  group_ratio,
  ci = 0.95
)

precision_rate_ratio(n_exposed, exposed, unexposed, group_ratio, ci = 0.95)

precision_odds_ratio(
  n_cases,
  exposed_cases,
  exposed_controls,
  group_ratio,
  ci = 0.95
)

Arguments

Value

a tibble with precision, effect measure, and sample size

References

Rothman, K.J. and Greenland, S. 2018. Planning Study Size Based on Precision Rather Than Power. 29(5):599-603.

Examples

# From Rothman and Greenland 2018

precision_odds_ratio(
  n_cases = 500,
  exposed_cases = .6,
  exposed_controls = .4,
  group_ratio = 2
)

Minimalist themes for precision plots

Description

Minimalist themes for precision plots

Usage

theme_precisely(base_size = 14, base_family = "", ...)

Arguments

Estimate sample size based on probability that upper limit is below level of concern.

Description

These functions calculate sample size based on probability that upper limit is below level of concern. The idea behind this approach is to use precision to provide support for the absence of effect. These functions calculate sample size where, when the true effect is null, the upper limit of the confidence interval of the estimate of interest has a probability of being at or under a specified level of concern.

Usage

upper_risk_difference(
  upper_limit,
  prob,
  exposed,
  unexposed,
  group_ratio,
  ci = 0.95
)

upper_risk_ratio(upper_limit, prob, exposed, unexposed, group_ratio, ci = 0.95)

upper_rate_difference(
  upper_limit,
  prob,
  exposed,
  unexposed,
  group_ratio,
  ci = 0.95
)

upper_rate_ratio(upper_limit, prob, exposed, unexposed, group_ratio, ci = 0.95)

upper_odds_ratio(
  upper_limit,
  prob,
  exposed_cases,
  exposed_controls,
  group_ratio,
  ci = 0.95
)

Arguments

Value

a tibble with sample size, effect measure, upper limit, and probability

References

Rothman, K.J. and Greenland, S. 2018. Planning Study Size Based on Precision Rather Than Power. 29(5):599-603.

Examples

# From Rothman and Greenland 2018

upper_rate_ratio(
  upper_limit = 2,
  prob = .90,
  exposed = .01,
  unexposed = .01,
  group_ratio = 1
)