11-16-2017 10:00 AM - edited 11-16-2017 10:01 AM
I have a data set which displays a "lognormal" shaped dependence of Y on X. The data is radioactive emission rate as a function of time. I see a rapid rise in the early stages followed by a slower decay over time.
The code I use is as follows:
proc glimmix data = QUERY_FOR_SUB1CT;
model rate = time / dist=lognormal link=log noscale solution;
run;
I found this code in an example in the communities. My question is how do I translate the results into a function that I can use in another program such as Excel. My parameter estimates are:
Parameter Estimates | |||||
Effect | Estimate | Standard | DF | t Value | Pr > |t| |
Intercept | -17.2289 | 65.7665 | 29249 | -0.26 | 0.7933 |
time | -0.00013 | 0 | 29249 | -Infty | <.0001 |
Scale | 11.9519 | 0.09883 | . | . | . |
I see on Wikipedia that the lognormal pdf is as attached in the pic but I fail to see how to populate this equation given the SAS output.
11-16-2017 10:23 AM
Are you sure that you want to use LINK=LOG?
To model lognormal data, most people use only DIST=LOGNORMAL and an identity link function.The assumption is that
LOG(rate) is normally distributed and is estimated by
LOG(RATE) = -17.23 - 0.00013*Time + eps
where eps (the residuals) is normally distributed with a variance that is estimated by the scale parameter.
11-16-2017 10:23 AM
Are you sure that you want to use LINK=LOG?
To model lognormal data, most people use only DIST=LOGNORMAL and an identity link function.The assumption is that
LOG(rate) is normally distributed and is estimated by
LOG(RATE) = -17.23 - 0.00013*Time + eps
where eps (the residuals) is normally distributed with a variance that is estimated by the scale parameter.
11-16-2017 10:29 AM
You are correct. I forgot to alter my code in what I posted above. On my machine I am using an identity link, such that results appear as follows.
Parameter Estimates | |||||
Effect | Estimate | Standard | DF | t Value | Pr > |t| |
Intercept | -2.7970 | 0.01497 | 29249 | -186.85 | <.0001 |
time | -0.00124 | 0.000120 | 29249 | -10.28 | <.0001 |
Scale | 3.5190 | 0.02910 | . | . | . |
Need further help from the community? Please ask a new question.