R vs SAS Confidence Intervals for Independent Proportions

Introduction

See the summary page for general introductory information on confidence intervals for proportions, including the principles underlying the most common methods.

For more technical derivation of methods for comparing independent proportions, see the corresponding SAS page.

The tables below provide an overview of findings from R & SAS, for calculation of CIs for comparisons of 2 independent proportions.

Note that if the number of responses in both groups is zero, SAS fails to produce confidence intervals for the difference.

Proportion Difference

Analysis of Two Independent Sample Proportions	Supported in R	Supported in SAS	Results Match
Normal approximation (Wald Method)	Yes {ratesci} `rdci()` Yes {DescTools} `BinomDiffCI(..,method = c("wald"))`	Yes (default)	Yes and results match by hand calculation
Agresti-Caffo	Yes {ratesci} `rdci()` Yes {DescTools} `BinomDiffCI(..,method=c("ac"))`	Yes	Yes
MOVER Wilson (Newcombe hybrid score) method	Yes {ratesci} `rdci()` Yes {DescTools} `BinomDiffCI(..,method=c("score"))`	Yes (‘Newcombe’)	Yes and results match by hand calculation. Warning: {DescTools} ‘method=score’ not to be confused with Asymptotic Score methods.
MOVER Jeffreys	Yes {ratesci} `rdci()`	No	NA
Miettinen-Nurminen Asymptotic Score	Yes {ratesci} `rdci()` Yes {DescTools} `BinomDiffCI(..,method=c("mn"))`	Yes	Yes
Mee Asymptotic Score	Yes {ratesci} `rdci()` Yes {DescTools} `BinomDiffCI(..,method=c("mn"))`	Yes (‘Miettinen-Nurminen-Mee’)	Yes
Skewness-corrected Asymptotic Score	Yes {ratesci} `rdci()` Yes {ratesci} `scasci(..,contrast="RD")`	No	NA

Normal approximation (Wald Method) with continuity correction	Yes {ratesci} `rdci(..., cc=TRUE)` Yes {DescTools} `BinomDiffCI(..,method=c("waldcc"))`	Yes	Yes
Hauck-Anderson continuity adjusted	Yes {ratesci} `rdci(..., cc=TRUE)` Yes {DescTools} `BinomDiffCI(..,method=c("ha"))`	Yes	Yes
MOVER Wilson (Newcombe) method with continuity adjustment	Yes {ratesci} `rdci(..., cc=TRUE)` Yes {DescTools} `BinomDiffCI(..,method=c("scorecc"))`	Yes (‘Newcombe (Corrected)’)	Yes
MOVER Jeffreys with continuity adjustment	Yes {ratesci} `rdci(..., cc=TRUE)`	No	NA
Asymptotic Score methods with continuity adjustment	Yes {ratesci} `rdci(..., cc=TRUE)`	No	NA
‘Exact’ methods	No	Yes	NA

Relative Risk

(Continuity-adjusted methods omitted for brevity)

Analysis of Two Independent Sample Proportions	Supported in R	Supported in SAS	Results Match
Normal approximation (Wald/Katz log Method)	Yes {ratesci} `rrci()` Yes {DescTools} `BinomRatioCI(..,method=c("katz.log"))`	Yes (default)	Yes
Likelihood ratio	No	Yes `CL=LR`	NA
MOVER-R Wilson	Yes {ratesci} `rrci()`	No	NA
MOVER-R Jeffreys	Yes {ratesci} `rrci()`	No	NA
Miettinen-Nurminen Asymptotic Score	Yes {ratesci} `rrci()` Yes {DescTools} `BinomRatioCI(..,method=c("mn"))`	Yes (‘Score’) `CL=score`	Yes
Koopman Asymptotic Score	Yes {ratesci} `rrci()` Yes {DescTools} `BinomDiffCI(..,method=c("mn"))`	Yes `CL=(score(correct=no))`	Yes
Skewness-corrected Asymptotic Score	Yes {ratesci} `rrci()`	No	NA
Continuity adjusted methods	Yes {ratesci} `rrci(..., cc=TRUE)`	No	NA

Odds Ratio

(Continuity-adjusted methods omitted for brevity)

Analysis of Two Independent Sample Proportions	Supported in R	Supported in SAS	Results Match
Normal approximation (Wald/Woolf logit Method)	Yes {ratesci} `orci()` Yes {contingencytables} `Woolf_logit_CI_2x2()`	Yes (default)	Yes
Likelihood ratio	No	Yes `CL=LR`	NA
Mid-P	Yes {contingencytables} `Cornfield_midP_CI_2x2()`	Yes	Yes
MOVER-R Wilson	Yes {ratesci} `orci()`	No	NA
MOVER-R Jeffreys	Yes {ratesci} `orci()`	No	NA
Miettinen-Nurminen Asymptotic Score	Yes {ratesci} `orci()`	Yes (‘Score’) `CL=score`	Yes
Uncorrected Asymptotic Score	Yes {ratesci} `orci()`	Yes `CL=(score(correct=no))`	Yes
Skewness-corrected Asymptotic Score	Yes {ratesci} `orci()`	No	NA
Continuity adjusted methods	Yes {ratesci} `orci(..., cc=TRUE)`	No	NA

Prerequisites: R Packages

See the R page for more detail.

# Example R packages required
# pak::pak("petelaud/ratesci") # development version from GitHub
library(ratesci)