prtrend is the p-value associated with the 'recent trend' - typically the trend in the last 20 years
A pathological case occurs when:
a linear model has been fitted (normal errors)
all the data in the recent period (apart from the very first year) are censored
In this case, the fitted values are all the same, the recent trend is estimated to be zero, and yet prtrend is highly significant. This arises because prtrend is pickup from p_linear. In this case, it makes sense to obtain prtrend from the contrasts, which will give a value of unity (i.e. no evidence of any trend).
There is a more general issue here, related to the definition of recent trends. This can be dealt with in the release after next. I'll raise an issue.
prtrendis the p-value associated with the 'recent trend' - typically the trend in the last 20 yearsA pathological case occurs when:
In this case, the fitted values are all the same, the recent trend is estimated to be zero, and yet
prtrendis highly significant. This arises becauseprtrendis pickup fromp_linear. In this case, it makes sense to obtainprtrendfrom the contrasts, which will give a value of unity (i.e. no evidence of any trend).There is a more general issue here, related to the definition of recent trends. This can be dealt with in the release after next. I'll raise an issue.