@KsharpOne more question. When solving for Cr, why is sqrt followed by "&n1." instead of "&n.".   ... View more

@Ksharp  hi Ksharp,long time no see. I ran the pb regression program again in the past few days and found a problem. The temp3 dataset needs to delete the sij=-1 observation, but I opened the dataset and still saw the observation with sij=-1.See figure below. Then I ran the program to see the number of observations sij=-1 in the temp2 and temp3 datasets, and the results were strange.The results show that there is data for sij=-1 in temp2, but no data for sij=-1 in temp3.See figure below. But this is obviously inconsistent with the dataset results, which should have an impact on the results of the program, right? Why is this happening? ... View more

Thanks to your reply. I learned a lot from our disscution. ... View more

There is also a small question, why is the same 1000 sampling times, SAS running much slower than MedCalc? ... View more

Thanks for your reply and discussion, after looking at the data today and verifying it with another set of data, I have the following ideas: 1.In method comparison, x and y are usually not much different values and are often used for equivalence analysis.For example, the concentration of glucose in serum is tested with 2 IVD products, so that x and y are very close. 2.The data you used to test your code before, x and y, are very different, which in my opinion is already a significant inconsistency 3.So, this could be the reason why the results are problematic? Maybe MedCalc has its own way of handling similar data? 4.I think (of course, my experience is not enough), the main application area of PB regression is the analysis of laboratory test results,which is regression analysis between two similar data. So this problem is still very important and needs to be solved. 5.Another way of thinking, MedCalc has a built-in way of processing very small data, because many of the previous sets of data are 0.0X, or maybe this causes a slight deviation in the results? 6.I used another set of real data to verify, and the results are as follows(with Bootstrap CI😞 MedCalc: SAS:   data:   data have; label x="micrograms per deciliter" y="kiloOhms"; input x y @@; datalines; 3.28 3.42 1.91 2.22 15.93 17.45 16.72 17.16 13.83 14.38 1.81 1.94 9.61 8.43 16.06 17.69 16.46 18.10 5.75 7.53 0.053 0.06 17.75 20.28 13.72 15.30 0.857 1.01 14.63 16.08 16.23 15.51 4.42 4.87 2.02 2.33 17.74 19.18 4.12 3.12 15.90 17.16 4.89 5.36 1.65 1.56 21.37 22.71 0.296 0.33 2.66 2.68 11.46 12.07 18.16 16.88 1.85 1.73 11.83 13.61 0.998 1.24 3.80 4.22 6.79 7.85 15.37 18.06 1.37 1.48 8.11 9.88 0.145 0.16 1.63 1.61 6.15 5.88 5.80 5.76 2.41 2.39 17.03 17.32 0.343 0.40 3.75 3.89 8.05 8.95 7.38 6.82 9.76 11.29 3.43 4.24 7.07 8.64 8.46 8.97 6.25 7.24 1.66 1.36 3.98 4.21 9.11 8.08 10.83 11.47 4.66 4.43 8.90 7.89 15.43 15.60 2.87 2.33 0.032 0.03 19.80 21.17 14.92 15.83 6.14 6.68 3.90 4.20 0.132 0.14 0.824 0.92 4.09 4.63 3.22 3.39 7.56 8.13 5.67 6.96 1.53 1.07 5.21 6.58 9.66 8.57 6.48 6.44 1.99 2.29 21.96 22.77 18.08 18.45 11.89 13.24 11.65 11.97 15.02 15.71 0.440 0.42 19.12 20.80 7.13 7.17 16.15 17.54 15.91 17.05 17.76 18.62 11.81 12.12 5.34 5.00 6.19 7.55 5.59 5.66 9.39 9.35 9.45 9.24 6.14 6.67 4.30 4.83 11.46 10.35 3.99 4.34 3.51 3.53 3.33 3.18 9.94 8.58 12.77 14.25 1.56 1.61 11.10 10.56 2.20 2.60 6.12 6.64 8.90 9.79 12.46 14.08 10.05 10.10 0.053 0.05 4.22 3.89 5.76 6.78 1.54 1.06 13.68 14.65 3.42 3.44 2.03 2.46 9.91 12.66 0.832 0.93 2.40 1.88 4.99 6.20 5.90 6.69 0.527 0.59 0.404 0.51 0.450 0.41 0.146 0.15 0.040 0.04 1.23 1.43 5.68 6.31 0.231 0.27 0.254 0.26 0.346 0.38 0.239 0.27 0.311 0.27 3.30 3.22 14.22 13.28 14.18 15.75 0.621 0.53 11.75 12.65 10.32 9.34 15.48 16.43 8.93 7.86 4.05 4.23 11.18 12.65 14.60 15.85 3.76 3.23 9.55 8.99 12.06 11.54 2.00 1.69 11.88 12.07 7.17 7.53 14.54 14.75 5.47 6.34 5.70 5.03 9.69 9.43 0.034 0.03 9.50 9.03 10.35 9.54 0.468 0.44 14.69 15.86 2.57 2.15 1.29 1.22 2.12 1.98 0.061 0.06 12.52 14.54 10.29 12.04 3.42 2.99 1.94 2.04 1.06 0.98 4.61 4.23 3.49 3.22 12.52 14.35 1.29 1.32 6.18 7.12 ; run; ... View more

Thanks for your reply. @Ksharp To verify the accuracy of the code more comprehensively, I used a different dataset to verify it. But different results have been obtained. The point estimates for slope and intercept have been slightly different. In my opinion, the point estimate does not involve a random sampling algorithm and should be the same, so is there any factor that is not reflected in the code? SAS result： MedCalc result： My data: data have; label x="micrograms per deciliter" y="kiloOhms"; input x y @@; datalines; 0.07 0.07 0.04 0.10 0.07 0.08 0.05 0.10 0.06 0.08 0.06 0.07 0.03 0.05 0.04 0.05 0.05 0.08 0.05 0.10 0.12 0.21 0.10 0.17 0.03 0.11 0.18 0.24 0.29 0.33 0.05 0.05 0.04 0.05 0.11 0.21 0.38 0.36 0.04 0.05 0.03 0.06 0.06 0.07 0.46 0.27 0.08 0.05 0.02 0.09 0.12 0.23 0.16 0.25 0.07 0.10 0.07 0.07 0.04 0.06 0.11 0.23 0.23 0.21 0.15 0.20 0.05 0.07 0.04 0.14 0.05 0.11 0.17 0.26 8.43 8.12 1.86 1.47 0.17 0.27 8.29 7.71 5.98 5.26 0.57 0.43 0.07 0.10 0.06 0.11 0.21 0.24 8.29 7.84 0.28 0.23 8.25 7.79 0.18 0.25 0.17 0.22 0.15 0.21 8.33 7.67 5.58 5.05 0.09 0.14 1.00 3.72 0.09 0.13 3.88 3.54 0.51 0.42 4.09 3.64 0.89 0.71 5.61 5.18 4.52 4.15 0.09 0.12 0.05 0.05 0.05 0.05 0.04 0.07 0.05 0.07 0.09 0.11 0.03 0.05 2.27 2.08 1.50 1.21 5.05 4.46 0.22 0.25 2.13 1.93 0.05 0.08 4.09 3.61 1.46 1.13 1.20 0.97 0.02 0.05 1.00 1.00 3.39 3.11 1.00 0.05 2.07 1.83 6.68 6.06 3.00 2.97 0.06 0.09 7.17 6.55 1.00 1.00 2.00 0.05 2.91 2.45 3.92 3.36 1.00 1.00 7.20 6.88 6.42 5.83 2.38 2.04 1.97 1.76 4.72 4.37 1.64 1.25 5.48 4.77 5.54 5.16 0.02 0.06 ; run; ... View more

Thanks for your discussion of this issue. @Ksharp,@FreelanceReinhard  I used MedCalc version 20.026.This is a relatively new software version, and I can understand that @FreelanceReinhard your software version has not yet taken into account the estimation method of the confidence interval.In fact, medcalc already takes into account algorithm updates for confidence intervals. For confidence interval estimation for PB regression, the Bootstrap or Jackknife method is necessary and scientific. So, I would like to suggest to you, can you use these two algorithms to optimize your code? @KsharpThis is very important for practitioners in the IVD industry or in the field of medical testing. As IVD practitioners, and SAS users, it is very desirable to integrate these algorithms on THE SAS without the need to use other statistical software.I wonder what the advice of you two are? The figure below is medcalc default replocations and random-number seed.   ... View more

Hello, I read your code and found it enlightening. And verified it on Medcalc with your data, but found that the results were slightly different. But in my ability, I can't tell why the difference is happening, so here's a tip for you to continue your research and look forward to some more research on the code.   ... View more