Ranuni seeds and the proc reg procedure

[deleted_user]

I am confused with one problem that might due to ranuni function or the proc reg procedure. Here is the situation:
1. I created several dummy variables and try to create a new data with contains only the dummy variables and their source raw variables. Since the file is big, I split the file into two parts as my testing data and validation data. Here I use the ranuni to split.
2. In one of my procedure, I use proc reg and sas return a message indicating that my liner model is not fully ranked due to 'since the variables are a linear combination of other variables as shown'.

Since I am pretty sure that this dummy variable could not be the linear combination of others, so I went back and check my codes. Well, nothing went wrong. However, after I changed the seed of the ranuni, that warning message just disappeared and my codes worked fine.

Could anyone tell me why this happen? Does it relate to the mechansim that ranuni works in SAS?


Thank you very much for any comments!

[deleted_user]

creating dummy variables without any restriction can endup with linear dependency (once in thousands). its name is ranuni (random). better use a big seed number.