equiv.boot.lm {equivalence} | R Documentation |
This is called by equiv.boot(). This function fits an OLS regression to the observations denoted by index, assess whether the intercept and slope are within their respective regions of similarity, and returns the outcome as well as the estimates.
equiv.boot.lm(compare, index, rel.int.int = 0.25, rel.int.slope = 0.25)
compare |
a pair of equal length vectors, response first, predictor second. |
index |
index integers supplied by boot. |
rel.int.int |
the relative size of the region of similarity for the intercept. |
rel.int.slope |
the relative size of the region of similarity for the slope. |
The outcome is assessed via the bootstrap statistics, which are a four-row table. We focus on the first two, corresponding to the intercept region of similarity and the slope region of similarity. The original column shows whether or not the fitted line is within the interval (0 or 1). The bias column shows the difference between the first column and the bootstrap-estimated proportion. Obtain the bootstrapped statistics using apply (see the examples).
The estimates of slope and intercept are kept for interest, e.g. to check the distribution, obtain estimates of standard errors, etc.
The output is a four-component vector, comprising:
Boolean |
is the estimated intercept below the region of similarity? |
Boolean |
is the estimated intercept within the region of similarity? |
Boolean |
is the estimated intercept above the region of similarity? |
Boolean |
is the estimated slope below the region of similarity? |
Boolean |
is the estimated slope within the region of similarity? |
Boolean |
is the estimated slope above the region of similarity? |
Double |
estimated intercept |
Double |
estimated slope |
Andrew Robinson A.Robinson@ms.unimelb.edu.au
Robinson, A.P., R.A. Duursma, and J.D. Marshall. 2005. A regression-based equivalence test for model validation: shifting the burden of proof. Tree Physiology 25, 903-913.
lm
, boot
, equiv.boot
data(pref.4PG) FourPG.plot.boot <- boot(cbind(pref.4PG$stemvolinc, pref.4PG$volinc4PG), equiv.boot.lm, R=100, rel.int.int=0.25, rel.int.slope=0.25) FourPG.plot.boot apply(FourPG.plot.boot$t, 2, mean)[1:6] data(pref.LAI) lai.plot.boot <- boot(cbind(pref.LAI$lai.pa, pref.LAI$lai.bl), equiv.boot.lm, R=100, rel.int.int=0.25, rel.int.slope=0.25) lai.plot.boot apply(lai.plot.boot$t, 2, mean)[1:6] data(ufc) ufc.plot.boot <- boot(cbind(ufc$Height.m, ufc$Height.m.p), equiv.boot.lm, R=100, rel.int.int=0.25, rel.int.slope=0.25) ufc.plot.boot apply(ufc.plot.boot$t, 2, mean)[1:6]