14/0 14/02/ 2/20 2016 16
How can can I calcula calculate te "net "net reclassif reclassificat ication ion index index - NRI" using using SPSS SPSS or othe otherr statist statistica icall progra programs? ms? - Resea ResearchG rchGat ate e We use cookies to give you the best possible experience on ResearchGat ResearchGate. e. Read our cookies policy to learn more.
Ok
Abduzh bduzhappar appar Gaipov National Nation al Scientific Medical Scientific Medical Research Center, Astana, Kazakhs...
QUESTION FOLLOWERS (9)
How can I calculate "net reclassification index - NRI" using SPSS or other statistical programs? I should analyse some biomarkers predicting development of AKI using NRI. I have no idea how to do it via SPSS.
3100
Views
9
Followers TOPICS M ul ul titi va var ia te te S ta ta titi st st ic ics
S AS AS
S PS PS S
D at at a A na na lysi s
22
Answers
S ta ta titi st st ica l S of of tw tw ar ar e
Feb 2, 2014
Share
1
/1
ALL ANSWERS ANSWERS (22)
Pascal Nitiema · University of Oklahoma Health Sciences Center
1) You may take a look at the R packages: - ‘PredictABEL’, with the manual available at: http://cran.rproject.org/web/packages/PredictABEL/PredictABEL.pdf - ‘survIDINRI’, with its manual at: http:/ /cran /cr an.r.rproject.org/web/packages/survIDINRI/survIDINRI.pdf project. org/web/packages/survIDINRI/survIDINRI.pdf 2) For SAS see: http://analytics.ncsu.edu/sesug/2010/SDA07.Kennedy.pdf Hope this help. Cheers. Feb 4, 2014
Abduzhappar Gaipov · National Scientific Medical Research Center, Astana, Kazakhstan
Dear Dr Pascal Thank you very much I will try it Feb 5, 2014
John W Pickering · University of Otago
I've put some R and Matlab functions for NRI, IDI, category free NRI, and Risk Assessment Plots Plots on Research Gate. Attached below: Dataset: Matlab RAP Code John W Pickering · Zoltan H Endre [Show abstract] Full-text · Dataset · Jan 2013
Download Source
Dataset: R function for Risk Assessment Plot & reclassification metrics NRI, IDI, cfNRI
Got a ques question tion you need answered quickly? John W Pickering · Zoltan H Endre [Show abstract]
Sign up today to join our community of over 8 million scientific professionals.
Technical questions questions like the one you've just found usually Full-text · Dataset · Aug 2014
get answered within 48 hours on ResearchGate.
Join for free
14/0 14/02/ 2/20 2016 16
How can can I calcula calculate te "net "net reclassif reclassificat ication ion index index - NRI" using using SPSS SPSS or othe otherr statist statistica icall progra programs? ms? - Resea ResearchG rchGat ate e Aug 17, 2014
Gehendra Mahara · Capital Medical University
Dear Abduzhappar Gaipov Package ‘PredictABEL’ this is the PDF file you can find the detail about it. and also u can find it in this link. http://cran.r-project.org/web/packages/PredictABEL/PredictABEL.pdf Thank you. Good luck May 19, 2015
John W Pickering · University of Otago
You are welcome to use the attached excell spreadsheet for calculating NRI & IDI. Obviously it doesn't calculate the confidence interval, interval, which is the important metric, but I use it to check my matlab or R calculations are good. (ps how's the AKI research going?). NRI_IDI.xlsx May 19, 2015
Mariana Rodrigues · Serviço de Saúde da RAM, E.P.E.
Hello John W Pickering, i'm trying to understand how to calculate NRI throw the excel file and i have some doubts in the values of the models. How can i calculate those models in the spss software? Thank you. May 20, 2015
John W Pickering · University of Otago
Hola Mariana, I don't think SPSS can calculate NRI yet (I no longer use SPSS, so I'm not 100% sure). I too have doubts about the value of the models for NRI, particularly because of arbitrary choices of thresholds (I've attached a paper where I show why). The IDI is better (if reported for those with and without events separately). Sorry I can't help with SPSS. If you want to message me I am happy to discuss my running some data for you through my R program. John Article: New Metrics for Assessing Diagnostic Potential of Candidate Biomarkers John W Pickering · Zoltan H Endre [Show abstract]
Source
Full-text · Article · Jun 2012 · Clinical Journal of the American Society of Nephrology
Download May 20, 2015
Gehendra Mahara · Capital Medical University
Dear Dr John You are right that it is quite difficult t do on SPSS. It is easier to do with R software on PredictABEL packege. Thanks Gehendra May 21, 2015
Mariana Rodrigues · Serviço de Saúde da RAM, E.P.E.
14/0 14/02/ 2/20 2016 16
How can can I calcula calculate te "net "net reclassif reclassificat ication ion index index - NRI" using using SPSS SPSS or othe otherr statist statistica icall progra programs? ms? - Resea ResearchG rchGat ate e models were the hazard ratio in the survival analysis. Can anyone tell me if i'm right or tell me what those values means? Thanks Mariana Rodrigues May 21, 2015
Kirsten E Peters · University of Western Australia
Dear John (and others), I'm very interested in calculating NRI/IDI for my data. Can your excel file be used for logistic regression models, and are the data in columns B and C of your xls file predicted probabilities? I'm not good with R, but am trying to learn SAS... Thanks, Kirsten Jul 28, 2015
John W Pickering · University of Otago
Dear Kirsten, yes, the data in columns B & C are the predicted probabilities derived from logistic regression regression models. I'm sorry, but I don't use SAS, so can't help you with that. Perhaps someone has written a module for reclassification. reclassification. The confidence intervals are important, so you'll need to be able to compute these. Be aware of the inadequacies of NRI (I only use it if t here are accepted pre-defined thresholds). Jul 28, 2015
Kirsten E Peters · University of Western Australia
Thanks for your quick reply John. I've found some SAS macros for NRI/IDI on the web, so just need t o figure out how to use those to compare with your xls file. Are the NRI/IDI values for events/non-events in your xls file acceptable to quote in a manuscript? I realise I will have to work out CI's and maybe even p-values too. I wish there was a way to do it in SPSS! Jul 29, 2015
Kirsten E Peters · University of Western Australia
Here's the results I obtained from your xls, which I'm wondering how to interpret?
Events nup 15 20 35 NRIevents -14.28571429 NonEvents ndown 165 127 292 292 NRInon-events -13.01369863 Events sum risk up 1.81379 -2.93219 35 IDIevents -0.031954286 NonEvents sum risk down 12.82108 -9.41447 292 IDInonevents -0.011666473 Jul 29, 2015
John W Pickering · University of Otago
Kirsten Of those with the event: 15 had an increase in calculated probability of the event, 20 a decrease. A net 14.3% had a decrease in risk (probably the "wrong" direction). The average change in risk was 0.032 (decrease). Of those without the event: 165 had an increase in calc probability, probability, 127 had a decrease, A net 13.0% had an increase in risk (probably in the "wrong" direction). direction). The average change in risk was 0.012 (increase). It seems that your "new model" is worse than the old model in that it decreased (on average) the calculated probability of the event for t hose that had the event, and increased it for those that didn't. If this is totally unexpected, check you have entered in Model 1 your reference model,
14/0 14/02/ 2/20 2016 16
How can can I calcula calculate te "net "net reclassif reclassificat ication ion index index - NRI" using using SPSS SPSS or othe otherr statist statistica icall progra programs? ms? - Resea ResearchG rchGat ate e Jul 29, 2015
Kirsten E Peters · University of Western Australia
Thanks John, that is what I feared! The AUC for model 2 is better t han model 1, so I must have copied my pred probs from the wrong models into the xls. I will check. YES, I copied the wrong model across...here's the update results which look more like it! Events nup 24 11 35 NRIevents 37.14285714 NonEvents ndown 112 180 292 NRInon-events 23.28767123 Events sum risk up 2.89515 -0.87759 35 IDIevents 0.057644571 NonEvents sum risk down 8.14748 -7.87663 292 IDInonevents -0.000927568 Jul 29, 2015
Kirsten E Peters · University of Western Australia
John, my conclusions now, based on your previous reply are: Of those with the event: 24 had an increase in calculated probability of the event, 11 a decrease. A net 37.1% had an increase in risk. The average change in risk was 0.058 (increase). Of those without the event: 112 had an increase in calculated probability, probability, 180 had a decrease. A net 23.3% had an decrease in risk. The average change in risk was 0.0009 (decrease). This shows that the "new model" is better than the old model in that it increased (on average) the calculated probability of the event for those that had the event, and decreased it for those that didn't. Is this correct? Jul 29, 2015
John W Pickering · University of Otago
Hi Kirstin Almost. 0.058 for the events is positive and and so in the "right" direction. Whether you can conclude an improvement depends on the confidence interval. the -0.0009 is negative (wrong direction) meaning that amongst those without the event an average average increase rather than decrease. decrease. However, I'm sure you'll find the confidence intervals will be such that they cross zero. ie the new model made no difference to those with the event. I think the NRIs are a waste of time for most assessments. I expect if you were to plot a Risk Assessment Plot you would see that the curve of for those without the events has a very small area under the curve - which is appropriate and suggestive there is little room f or improvement which it comes to assigning low risk to those who don't have the event. I suggest you message me if you would like to plots some RAPs & calc CIs. Jul 29, 2015
Thais Coutinho · University of Ottawa
Dear John: Thank you for the Xcel file above. I used it for my own data, and got the following output: Events nup 84
48 132 NRIevents 27.2727273
NonEvents ndown 330 141 471 NRInon-events -40.127389
14/0 14/02/ 2/20 2016 16
How can can I calcula calculate te "net "net reclassif reclassificat ication ion index index - NRI" using using SPSS SPSS or othe otherr statist statistica icall progra programs? ms? - Resea ResearchG rchGat ate e 0.00046006 NonEvents sum risk down 3.06411448 -1.5098106 471 IDInonevents -0.0033 Can you please help me interpret this output? Thank you! Thais 12 days ago
Thais Coutinho · University of Ottawa
Ah and one more question just to be sure: Model 1 is before the new predictor, and model 2 is after the new predictor, correct? 12 days ago
John W Pickering · University of Otago
Yes, that's correct Thais 12 days ago
John W Pickering · University of Otago
Thais - did my answer to your question about t he output from the file appear on the thread for you. you. It doesn't seem to appear for me now? 12 days ago
Thais Coutinho · University of Ottawa
OK, thank you John. So if I understood correctly, the new model was worse than the original one..? 12 days ago
Can you help by adding an answer?
Enter your answer
Add your answer
© 2008‐20 2016 16 re rese sear arch chga gate te.n .net et.. Al Alll rights g hts re rese serv rved ed..
Abou Ab outt us · Cont Contac actt us · Care Career erss · Deve Develo lope pers rs · News News · Help Help Ce Cent nter er · Priv Privac acyy · Term Termss | Ad Adve vert rtis isin ing g · Rec Recru ruit itin ing g