BestSlope: Choose the best-fit slope for the log(y) and x regression by...

View source: R/BestSlope.R

BestSlopeR Documentation

Choose the best-fit slope for the log(y) and x regression by the criteria of adjusted R-square.

Description

It sequentially fits (log(y) ~ x) from the last point of x to the previous points with at least 3 points. It chooses a slope the highest adjusted R-square. If the difference is less then 1e-4, it pickes longer slope.

Usage

  BestSlope(x, y, adm = "Extravascular", TOL=1e-4, excludeDelta = 1)

Arguments

x

vector values of x-axis, usually time

y

vector values of y-axis, usually concentration

adm

one of "Bolus" or "Infusion" or "Extravascular" to indicate drug administration mode

TOL

tolerance. See Phoneix WinNonlin 6.4 User's Guide p33 for the detail.

excludeDelta

Improvement of R2ADJ larger than this value could exclude the last point. Default value 1 is for the compatibility with other software.

Details

Choosing the best terminal slope (y in log scale) in pharmacokinetic analysis is somewhat challenging, and it could vary by analysis performer. Pheonix WinNonlin chooses a slope with highest adjusted R-squared and the longest one. The difference of adjusted R-Squared less than TOL considered to be 0. This function uses ordinary least square method (OLS). Author recommends to use excludeDelta option with about 0.3.

Value

R2

R-squared

R2ADJ

adjusted R-squared

LAMZNPT

number of points used for slope

LAMZ

negative of the slope, lambda_z

b0

intercept of the regression line

CORRXY

correlation of log(y) and x

LAMZLL

earliest x for lambda_z

LAMZUL

last x for lambda_z

CLSTP

predicted y value at the last point, predicted concentration for the last time point

Author(s)

Kyun-Seop Bae <k@acr.kr>

See Also

Slope

Examples

BestSlope(Theoph[Theoph$Subject==1, "Time"], Theoph[Theoph$Subject==1, "conc"])
BestSlope(Indometh[Indometh$Subject==1, "time"], Indometh[Indometh$Subject==1, "conc"],
          adm="Bolus")

NonCompart documentation built on Nov. 15, 2023, 9:06 a.m.