You could consider the CENSORED option, two-part modeling as shown in Example 6.16, or the method described in the following link on the website:
GLM with log link and Poisson regression for continuous variables
You might also want to get rid of the high values if they are outliers.
Robert Long posted on Sunday, October 20, 2013 - 6:47 am
Thank you for the response. I am now trying to implement a two-part model. Using example 6.16, could you tell me why the mean of the intercept growth factor for the binary part is fixed at zero, and whether this can be relaxed and if so, how ? Also, could you explain what the "Thresholds" part of the model results means with the four values of -0.541