I am trying to impute left-censored data by a multiple imputation. So left-censored data are missing.
I use PROC MI with method MCMC and use the option MAXIMUM=(value of censure) for that the imputed data to be below the censorhip threshold. But, I get an error message : "ERROR: An imputed variable value is not in the specified range after 100 tries."
I want impute the variable LB_LACTO_OLDnumConvertv1log.
Here is my code :
proc mi data=fusion seed=1234 nimpute=5 maximum=4.48 out=mimpute; mcmc; var LB_LACTO_OLDnumConvertv1log produit ; run;
If I don't use the option MAXIMUM= , the imputed data are above the threshold. Do you have a solution?
You have two variables listed on the VAR statement. The maximum statement you have will use that value as the maximum imputed value for all variables listed on the VAR statement. Is that what you want? If not, specify two maximum values on the maximum statement;
if you don't want a maximum for either of the two variables, use a period:
Thanks for your answer. But even if I put a point e.g "maximum = 4.48 . " , the values imputed are greater than the maximum 4.48. Maybe it's beacause all the other values observed are greater than 4.48 ? Would you know a solution ? Thanks.