From a publication, I have parameter estimates for A, B, and C in the following model : Yi = A * exp(-(Xi/B)**C)* ei
The publication also provides standard errors for those three estimates, the range of x values in the data from which the parameters were estimated, and nothing else.
I am generating predicted y values for some values of x within the range of the original data.
Can I also generate standard errors of those point predictions, and predicted percentiles?
Assuming the model was fit using regression assuming a normal response, you would need an estimate of the residual error. See the formula in "Statistical Background in Linear Regression / Predicted and Residual Values" in the "Statistical Background in Linear Regression" chapter at the beginning of the SAS/STAT User's Guide. Of course, your model is nonlinear, so see the formulas for LCL and UCL options in the "Details / Confidence intervals" section of the NLIN documentation. Even for a quantile-based interval, you would need an estimate of the normal variance.
Thank you for the response and for the references.
I read both, with only partial understanding.
Am I understanding your response and the references correctly that obtaining a standard error or percentiles for the point predictions is not possible given what is available to me?
Hi (and particularly @StatDave 😞
Would it be possible to generate a bootstrap estimate of the residual variance? Assume the range of X values is uniform. Assume the parameter estimates are normally distributed, with mean = the reported estimate and standard deviation = the reported standard error. Draw at random with replacement 100 values of X, and 1000 3-ples of the parameters (this is where a big assumption comes in - that the parameter estimates are uncorrelated). From these, calculate 100 x 1000 Y values. Last step, refit the equation at a range of (X,Y) values covering the interval, making sure to include X values of interest with Y set to missing. That should give predicted Y values with standard errors and confidence intervals, and I am sure quantiles could be calculated given the residual error.
Thoughts?
SteveDenham
Good news: We've extended SAS Hackathon registration until Sept. 12, so you still have time to be part of our biggest event yet – our five-year anniversary!
ANOVA, or Analysis Of Variance, is used to compare the averages or means of two or more populations to better understand how they differ. Watch this tutorial for more.
Find more tutorials on the SAS Users YouTube channel.