If I thought keeping track was absolutely necessary I would ensure that my original data has a unique identifier for each record. Then after I had checked/cleaned data I would have a data set of the identifiers checked. The "next time" I cleaned the data I would subset the data to those records whose idendifiers were not in the data set of the already checked. Then update the identifier set with those checked. Repeat as needed.   But there are a number of other issues involved I don't go into without getting paid... ... View more

If it was repeated measure ,then you should use NLMIXED , I will leave it to @Steave .   " are the p-value test for difference also correct on the original scale of the response variable?" Yes. I used ilink option, if you want see real mean value ,add option mean in it .     lsmeans member*group/ ilink exp diff cl mean ; ... View more

Run this program.   data op; input linear quadratic cubic; x + 1; cards ; -3 1 -1 -1 -1 3 1 -1 -3 3 1 1 ;;;; run; proc print; run; proc sgplot; pbspline x=x y=linear; pbspline x=x y=quadratic; pbspline x=x y=cubic; run;   ... View more

Not surprising that the "error" bounds are so wide--after all, it is a quadratic.  But that is why the delta method was worked out.  Here is a great source to get started (although the coding is in R):   http://www.ats.ucla.edu/stat/r/faq/deltamethod.htm   As far as interpreting, remember that the linear and quadratic terms are correlated, and so you must address both at once when looking at single-unit changes.  It is not like two separate regressors, where you can look at a unit change in one, and assume the other is unchanged.   Steve Denham ... View more

I think for Time series also smaller AIC and BIC is better. Could you please tell me why we need to choose the model with  smaller AIC and BIC? ... View more

Hi @Rick_SAS. Thank you for your reply. I have read but sadly I still don't understand. Please find my code below if that would be helpful. Thank you very much. ods graphics on; proc genmod data=temp plots=all; class country raps; model occamt=country raps country*raps / dist=gamma link=log type3; lsmeans raps country country*raps / pdiff ilink adj=tukey; contrast 'linear' raps -2 -1 0 1 2; contrast 'quadratic' raps 2 -1 -2 -1 2 ; contrast 'cubic' raps -1 2 0 -2 1; output out=Residuals pred=Pred resraw=Resraw reschi=Reschi resdev=Resdev stdreschi=Stdreschi stdresdev=Stdresdev reslik=Reslik; run; proc univariate data=Residuals normal; var Resraw Reschi Stdreschi Resdev Stdresdev Reslik; histogram/ normal; run; proc gplot data=residuals; plot Resdev*Pred; run; quit; proc gplot data=residuals; plot Stdresdev*Pred; run; quit; proc gplot data=residuals; plot Reslik*Pred; run; quit; proc gplot data=residuals; plot occamt*Pred; run; quit;   ... View more

y_pred is depended on Design Matrix , you missed the reference level of CLASS variable.   If I was right: Log(occ1)(When raps=1) = ...  0.5465*(raps=1).....    Log(occ0)(When raps=0 -- average effect of the four levels of raps )  = ...  0.5465*(raps=0).....   ==>     Log(occ-1/occ-0)= 0.5465   Which means E(log(occ)) will increase 0.5465 when raps turn from 0(average effect of the four levels of raps )  into 1(current level) . ... View more

Hi @Rick_SAS.  Thank you for your reply and all the links. I'm still trying to understand them. ... View more

In three=parameter families that have a threshold parameters, the data must be strictly larger than the values of the threshold parameter.  The reason is that the distributions you are fitting have a term that looks like log( x[i] - theta ) or (x[i] - theta)**c. The LOG function is undefined when its argument is zero or negative. Similarly, the "raise to power function" is undefined for negative arguments when  c is not an integer. (For example, c=1/2 is the square roo function.)   The most likely reason for these errors is that you have data values that are zero. Your choices include: Model the data by using different distributions Specify a value for the threshold parameters that is strictly less than the smallest data value Delete the bad data values if they are coding errors ... View more

Hi Norman. My apology for not making my problem clearer in the first place. Thank you for the link.   ... View more