contrast.rms collapses small p-values to zero about rms

[harrelfe]

To be consistent should that not be

|P <- if(length(idf)) 2 * pt(-abs(Z), idf)) else 2 * pnorm(-abs(Z)) |

On 09/30/2015 11:48 AM, Alexander Ploner wrote:

The function |contrast.rms| reports the p-values for all Wald
statistics absolutely larger than ca. 9 as an exact zero.

Example adapted from the |contrast.rms| help page with Z=10.36:

|set.seed(1) x = rnorm(500) xb = 1+2*x y = ifelse(runif(500) <=
plogis(xb), 1, 0) f = lrm(y ~ x) ct = contrast(f, list(x=1)) ct
ct$P==0 ## Exact zero |

The problem seems to be the line below in |contrast.s| where the
p-value is calculated:

|P <- if(length(idf)) 2 * (1 - pt(abs(Z), idf)) else 2 * (1 -
pnorm(abs(Z))) |

Replacing this with

|P <- if(length(idf)) 2 * (1 - pt(abs(Z), idf)) else 2 * pnorm(-abs(Z)) |

should do the trick.

|rms 4.3-1 _ platform x86_64-w64-mingw32 arch x86_64 os mingw32 system
x86_64, mingw32 status major 3 minor 2.2 year 2015 month 08 day 14 svn
rev 69053 language R version.string R version 3.2.2 (2015-08-14)
nickname Fire Safety |

—
Reply to this email directly or view it on GitHub
#13.

---

Frank E Harrell Jr Professor and Chairman School of Medicine

Department of *Biostatistics*   *Vanderbilt University*

[alexploner]

Of course! Sorry, completely blinded to dfs :) /alexander

On 2015-09-30 19:16, Frank Harrell wrote:

To be consistent should that not be

|P <- if(length(idf)) 2 * pt(-abs(Z), idf)) else 2 * pnorm(-abs(Z)) |

On 09/30/2015 11:48 AM, Alexander Ploner wrote:

The function |contrast.rms| reports the p-values for all Wald
statistics absolutely larger than ca. 9 as an exact zero.

Example adapted from the |contrast.rms| help page with Z=10.36:

|set.seed(1) x = rnorm(500) xb = 1+2*x y = ifelse(runif(500) <=
plogis(xb), 1, 0) f = lrm(y ~ x) ct = contrast(f, list(x=1)) ct
ct$P==0 ## Exact zero |

The problem seems to be the line below in |contrast.s| where the
p-value is calculated:

|P <- if(length(idf)) 2 * (1 - pt(abs(Z), idf)) else 2 * (1 -
pnorm(abs(Z))) |

Replacing this with

|P <- if(length(idf)) 2 * (1 - pt(abs(Z), idf)) else 2 *
pnorm(-abs(Z)) |

should do the trick.

|rms 4.3-1 _ platform x86_64-w64-mingw32 arch x86_64 os mingw32 system
x86_64, mingw32 status major 3 minor 2.2 year 2015 month 08 day 14 svn
rev 69053 language R version.string R version 3.2.2 (2015-08-14)
nickname Fire Safety |

—
Reply to this email directly or view it on GitHub
#13.

---

Frank E Harrell Jr Professor and Chairman School of Medicine

Department of Biostatistics Vanderbilt University

—
Reply to this email directly or view it on GitHub
#13 (comment).

Alexander Ploner
Medical Epidemiology & Biostatistics
Karolinska Institutet
http://ki.se/meb