Power analysis for detecting trends in linear regression is implemented following procedures in Gerrodette (1987; 1991).
powertrend(trend = 1, A1 = NULL, PSE = NULL, pserel = 1,
maxyrs = 3, pR = 100, step = 5, alpha = 0.05, tail = 2, graph = TRUE)1 = Linear, 2 = Exponential. Default = 1.
the start year abundance. In actuality, it can be population size, productivity, diversity, mortality rate, etc.
the proportional standard error (SE(A)/A) = CV in Gerrodette (1987;1991).
the relationship between abundance and PSE: 1 = 1/sqrt(A1), 2 = constant, 3 = sqrt(A1). Default = 1.
the maximum number of samples or years to project start year abundance. Default = 3.
the highest positive percent change to investigate. Default = 100.
the increment of the range of percent change to investigate. Default = 5.
the alpha level (Type I error) to use. Default = 0.05.
type of tailed test: 1 = one-tailed, 2= two-tailed. Default = 2.
logical specifying whether a graph of power versus percent change should be produced. Default is TRUE.
Dataframe containing columns of number of samples (years), trend selected (trend), the PSE (pse),
alpha level (alpha), tail of test (tail), percent change (R) over maxyrs, and power (power).
The probability that an upward or downward trend in abundance (power) will be detected is calculated using linear regression
given number of samples (maxyrs), estimates of sample variability (PSE) and abundance-PSE relationship (pserel),
and percent rate of change. The program calculates power for each step increment beginning at -100 percent for declining changes
and ending at pR percent for increasing changes. See Gerrodette (1987;1991)
for full details. It is assumed that time intervals between samplings is equal.
Gerrodette, T. 1987. A power analysis for detecting trends. Ecology. 68(5): 1364-1372.
Gerrodette, T. 1991. Models for power of detecting trends - a reply to Link and Hatfield. Ecology 72(5): 1889-1892.
# NOT RUN {
powertrend(A1=1000,PSE=0.1)
# }
Run the code above in your browser using DataLab