superb-package: superb: Summary Plots with Adjusted Error Bars

In dcousin3/superb: Summary Plots with Adjusted Error Bars

superb: Summary Plots with Adjusted Error Bars

Description

Computes standard error and confidence interval of various descriptive statistics under various designs and sampling schemes. The main function, superbPlot(), return a plot. superbData() returns a dataframe with the statistic and its precision interval so that other plotting package can be used. See Cousineau and colleagues (2021) \Sexpr[results=rd]{tools:::Rd_expr_doi("10.1177/25152459211035109")} or Cousineau (2017) \Sexpr[results=rd]{tools:::Rd_expr_doi("10.5709/acp-0214-z")} for a review as well as Cousineau (2005) \Sexpr[results=rd]{tools:::Rd_expr_doi("10.20982/tqmp.01.1.p042")}, Morey (2008) \Sexpr[results=rd]{tools:::Rd_expr_doi("10.20982/tqmp.04.2.p061")}, Baguley (2012) \Sexpr[results=rd]{tools:::Rd_expr_doi("10.3758/s13428-011-0123-7")}, Cousineau & Laurencelle (2016) \Sexpr[results=rd]{tools:::Rd_expr_doi("10.1037/met0000055")}, Cousineau & O'Brien (2014) \Sexpr[results=rd]{tools:::Rd_expr_doi("10.3758/s13428-013-0441-z")}, Calderini & Harding \Sexpr[results=rd]{tools:::Rd_expr_doi("10.20982/tqmp.15.1.p001")} for specific references.

Details

'suberb' is a library to perform descriptive statistics plots based on the superb framework. In a nutshell, the framework assert that confidence intervals must be devised according to all the relevant information that can be used to assess precision. For example, confidence intervals should be informed of the presence of within-subject design, of the fact that the sample is random or clustered, of whether the population is finite or infinite, etc.

Would you do a t-test on independent groups when you know that the data are paired? Of course, not! Why use the classic "stand-alone" confidence interval then? These classic confidence intervals are oblivious to most relevant information.

The superb framework is based on the idea that correct, well-informed, confidence intervals can be obtained with a succession of simple corrections. I call these "adjusted confidence intervals".

The main function is

superb(formula, dataframe, ...)

where df is a dataframe.

For more details on the underlying math, see \insertCitec05,c19,c17,cl16,m08,b12,lm94,gc19superb

A second function inserted in this package is \insertCitech19superb

GRD( ...)

which generates random datasets. It easily generate ficticious dataset so that superbPlot can be tested rapidly. This function is described in \insertCitech19superb.

Author(s)

Maintainer: Denis Cousineau [e-mail]

Other contributors:

• Bradley Harding [e-mail] [contributor]

• Marc-Andre Goulet [e-mail] [contributor]

• Jesika Walker [e-mail] [artist, presenter]

References

\insertAllCited

The package includes additional, helper, functions:

• ShroutFleissICC1 to compute intra-class correlation;

• epsilon to compute the sphericity measure;

• lambda to compute the cluster-sampling adjustment;

• MauchlySphericityTest to perform a test of sphericity;

• WinerCompoundSymmetry to perform a test of compound symmetry;

and example datasets described in the paper:

• dataFigure1 illustrate the paradox of using stand-alone CI in between-group design;

• dataFigure2 illustrate the paradox of using stand-alone CI in within-subject design;

• dataFigure3 illustrate the paradox of using stand-alone CI in cluster-randomized sampling study;

• dataFigure4 illustrate the paradox of using stand-alone CI with population of finite size.

See Also

Useful links:

• https://dcousin3.github.io/superb/

• Report bugs at https://github.com/dcousin3/superb/issues/

dcousin3/superb documentation built on Oct. 29, 2024, 5:28 p.m.