satpt

This R package identifies the saturation point of multinomial responses from a survey using standard errors of sample proportions.

Installation

You can install the current stable released version of satpt from CRAN with

install.packages("satpt")

when available on CRAN.

Development version

To get a bug fix or to use a feature from the development version, you can install the development version of satpt from GitHub.

# install.packages("remotes")
remotes::install_github("deboonstra/satpt")

Usage

For basic usage of satpt simply specify the responses of the survey in y and the data collection period information in by, as seen below.

library(satpt) # load package
data(diagnoses) # load example diagnoses data
res <- satpt::satpt(y = diagnoses$q2, by = diagnoses$wave)
print(res)

#> Analysis based on: q2 
#> Saturation achieved?  Yes 
#> 
#> Overall Sample Proportions and Standard Errors
#> ==============================================
#>             y: q2
#> Statistics   Not at all  Often   Once Rarely Sometimes
#>   Proportion     0.2531 0.0750 0.0375 0.3688    0.2656
#>   SE             0.0172 0.0104 0.0075 0.0191    0.0175

Saturation of each individual response category may be examined graphically while comparing the standard errors to the saturation threshold.

graphics::par(oma = c(0, 0, 0, 8))
plot(res)
# adding legend
satpt::legend_right(
  legend = "Saturation\nthreshold",
  col = "firebrick", lty = 3, lwd = 2,
  cex = 0.75
)

Methodology

Presented below is a simplified version of the algorithm that is employed in satpt to determine whether saturation of the responses has been achieved and whether pooled standard errors should be calculated to account for response bias.

Learn more

To get started, first read the Getting started with satpt vignette. Then, read more about how satpt may handle Select All Apply questions in the Impelementing with select all apply questions vignette.