Thanks a lot. I'll study the program. Great suggestions here to keep in mind. ... View more

@Reeza Reeza Thanks a lot. Trusted Advosor's suggestion made the standard program fit to my data better. Below code is (after suggestion applied) to score the questionnaire responses using 0-100 scale. I need someone to reassure with me on the program to be sure. Would you please look at it especially beyong /*Attention*/ mark? It makes sense to me, however, resulting data has many negative values (see attached csv).     DATA WANT; set HAVE; ARRAY PFI(10) PF01-PF10; DO I = 1 TO 10; PFMEAN = MEAN(OF PFI(I)); IF PFI(I)= . THEN PFI(I) = PFMEAN; END; /*Attention*/ PFNUM = N(OF PF01-PF10); IF PFNUM GE 5 THEN RAWPF = SUM(OF PF01-PF10); PF = ((RAWPF - 10)/(30-10)) * 100; LABEL PF = 'PF SCORING (0-100)' RAWPF = 'PF ITEM SUMMARY'; RUN;   ... View more

@PaigeMiller   Below is what I would do. Any objections? Here I am scoring data to 0-100 scale. So then the next step is to weight them.   DATA WANT; set HAVE; ARRAY PFI(10) PF01-PF10; DO I = 1 TO 10; PFMEAN = MEAN(OF PFI(I)); IF PFI(I)= . THEN PFI(I) = PFMEAN; END; PFNUM = N(OF PF01-PF10); IF PFNUM GE 5 THEN RAWPF = SUM(OF PF01-PF10); PF = ((RAWPF - 10)/(30-10)) * 100; LABEL PF = 'PF SCORING (0-100)' RAWPF = 'PF ITEM SUMMARY'; RUN;   ... View more

Hi Reeza, missing imputation and scoring program was provided to me by the author of the methodology. And i was trying to fit my data to the program provided. Now I'm willing to modify the program to the data, not other way around. ... View more

@PaigeMiller Very interesting. Zeros indicate "no, my physical activity is not limited" so is not missing. Assigning the column mean (mean(PF:)) would reflect in all study subjects' average versus imputing one's own responses' MEAN(OF PF01-PF10)?  Otherwise, average of same item(question) mean (PF01) instead taking average of different questions mean(PF01-PF10) makes sense to me.    ... View more

@PaigeMiller I'm dealing with survey data coded inconsistently. Shifting values by one unit to the right helped 7 out of 8 items included zero in the response choices. Now all zeros in the data is 1 so that those are not chopped off when following program is run. However, now my problem is that, one of items doesn't take 0 but starts from 1. After shifting, its starting value becomes 2.  I was wondering if it makes sense to shift 7 items to the right except that one item with no zero value anyway? Do you have any suggestions? thoughts?     ARRAY PFI(10) PF01-PF10; DO I = 1 TO 10; IF PFI(I) < 1 OR PFI(I) > 3 THEN PFI(I) = .; END; PFNUM = N(OF PF01-PF10); PFMEAN = MEAN(OF PF01-PF10); DO I = 1 TO 10; IF PFI(I)= . THEN PFI(I) = PFMEAN; END; IF PFNUM GE 5 THEN RAWPF = SUM(OF PF01-PF10); PF = ((RAWPF - 10)/(30-10)) * 100;   ... View more

@PatrickIt worked out except resulting missing in variable "subj"? strange? ... View more

Hi,   Final check, if you will. Below is how I'm approaching to negbin cdf plotting. Is it about right? Thanks much! Please ignore my previous question.   My reference: http://support.sas.com/kb/24/166.html   ESTIMATING OF NEGBIN PARAMETERS: proc genmod data=MYDATA; model outcome = / dist=negbin; ods output parameterestimates=post.pe; run; proc transpose data=post.pe out=post.tpe; var estimate; id parameter; run; data post.tpe; set; nb_k = 1/dispersion; nb_p = 1/(1+exp(intercept)*dispersion); nb_mean=nb_k*(1-nb_p)/nb_p; nb_var =nb_k*(1-nb_p)/nb_p**2; run; proc print; run; CALCULATING CUM PROBS: ods select cdfplot; proc univariate data=post.MYDATA; var rate; cdfplot rate / vscale=proportion vscale=proportion odstitle="Empirical CDF" odstitle2="PROC UNIVARIATE"; ods output cdfplot=post.outCDF; run; FEEDING NEGBIN CDF DISTRIBUTION WITH PARAMETERS CALCULATED ABOVE: data ALL; set outCDF; NBCDF = cdf("NegBin", ECDFX,0.027774,0.474); run; PLOTTING: title "MODEL FIT AFTER DEDUPLICATION"; proc sgplot data=ALL; step x=ECDFX y=ECDFY / legendlabel="ECDF"; step x=ECDFX y=NBCDF / legendlabel="Model Fit"; xaxis grid label="x" offsetmin=0.05 offsetmax=0.05; yaxis grid min=0 label="Cumulative Proportion"; run; ... View more

@Rick_SAS, @Reeza   Thanks Rick, i had updated the x axis scale with truncation syntax you suggested. Looking at the plots, I realized that negbin fits better than poisson not just slightly but significantly. Lambda estimated from my negbin model in proc genmod is 10.7 instead 1.04.   Can you give me more direct hints as to what x, 0.3 and 1 (as used in your previous demo) would be in the context of my data? I posted my data here for reference.   data mydata; set outCDF; NegbCDF = cdf("NegBin",x,0.3,1); run;       ... View more

you mean, take empirical cdf off of the plot for more constrast? ... View more

@Rick_SAS, Getting running out of time and steep learning curve, tempted to stick to poisson dist. defintely shall put more efforts in to understanding negbin later as it offers slightly a better fit to my data. Looking at below cdf plots, my data cleaning seemed not to cause an appreciable difference. I don't know If I could even tell the plot on the right (model fit after deduplication)  shows a better fit. What do you think? Please let me know if there're better diagnostic visuals that would show the impact of data cleaning on model fitting with more contrast than this?     ... View more

@Rick_SAS, Thanks for previous comments. All worked out.   1. I tried following but NegBCDF column is all missing in the outCDFcar. Lambda1 is the exp(estimate) from "ods output parameterestimates=data;"      ods select cdfplot; proc univariate data=sashelp.cars; cdfplot mpg_city / vscale=proportion odstitle="Empirical CDF" odstitle2="PROC UNIVARIATE"; ods output cdfplot=outCDFcar; run; %let Lambda1=1.04; data alldup; set outCDFcar; NegBCDF=cdf('NEGB',1, ECDFX, &Lambda1); run; 2. Below is the desired image. I have two separate cdf data with N=1122 for clean and N=1133 for uncleaned data from proc univariate. Any idea how I can overlay them on the same plot as shown below?      ... View more