glmReserve

[grishaloshadinov]

when I run "glmReserve" in RExcel with "var.power=1" and "cum=FALSE" and "mse.method = bootstarp" and "nsim =1000" I get error message "Microsoft Excel is waiting for another application to complete an OLE action"
When I run it in R it takes absolutely ages and eventually I have to kill it with getting no results

[trinostics]

It would appear that your RExcel issue and your "R it takes absolutely ages" issue are related. Please provide a minimal, reproducible example without RExcel. If you cannot reproduce the issue without RExcel, please resolve your issue with the RExcel people.

[grishaloshadinov]

I can't reproduce it in R either. How can I attach excel file with raw triangle that I am using please?

[Rick132]

Hello!
I have the same problem. I am working in R. When I run it in R it takes absolutely ages and eventually I have to kill it with getting no results. I hope someone can help

[actuaryzhang]

Is there an example you can provide to reproduce the issue? Is this issue with glmReserve or RExcel?

[grishaloshadinov]

that's what I mean, they didn't take me seriously, more flaws will follow

On Monday, 25 January 2016, 17:12, actuaryzhang <[email protected]> wrote:

Is there an example you can provide to reproduce the issue? Is this issue with glmReserve or RExcel? —
Reply to this email directly or view it on GitHub.

[Rick132]

Hello there!
Thanks for replying. I am using glmReserve in R. Find attached the file. I am not sure if it because the incremental payment contains some zeros. Also, is there a way I can extract the process and parameter risk from glmReserve either with analytical or bootstrapping? Do the package provide RMSEP of Munich Chain Ladder? Hoping to get a response soon.
glmreserve1.pdf

[actuaryzhang]

Thank you very much for providing the example. It seems that the resampling generates lots of negative values of the incrementals, which are not allowed in the current implementation. The reason that it takes forever is because it keeps resampling until all values are nonnegative. This criterion is an intuitive but perhaps not the best way to handle negative payments. I will investigate this problem more and try to use a better approach. Hopefully I can solve this soon but no promise given my current schedule. I will keep you posted.

Right now, the glmReserve function does not report separately the process and estimation errors. It is available, however, in the code. I can make the function report these details in the next release, but I seldom find such a breakdown meaningful.

[Rick132]

Thanks for the reply.

[kennedymwavu]

Has anyone found workarounds for this issue? I'm facing the same problem in R.

[trinostics]

@actuaryzhang Hi Wayne. Is it possible to modify the procedure to use real valued residuals? Long time no see. Dan

[actuaryzhang]

Hi Dan, good to hear from you. It's been a long time since the last time I worked on this. IIRC, the issue is fitting the GLM on data with negative values generated from the simulation. One workaround is setting negative values to zero (and redistributing the value to other cells). Feel free to make this change. Wayne

…

On Wed, Mar 16, 2022 at 1:18 PM Dan Murphy ***@***.***> wrote: @actuaryzhang <https://github.com/actuaryzhang> Hi Wayne. Is it possible to modify the procedure to use real valued residuals? Long time no see. Dan — Reply to this email directly, view it on GitHub <#2 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACURVADG5ME6K3A4SACFBLTVAI6YRANCNFSM4A3MNPFA> . Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>. You are receiving this because you were mentioned.Message ID: ***@***.***>

[trinostics]

@actuaryzhang In another context I have chosen to "wipe out" negative simulated values when a subsequent dependent calculation / function requires positive support (like 'log'). Is this a similar situation? What about replacing the negative values with their absolute values, rather than with zero? If replace with zero, what do you mean by "redistributing the value to other cells"?

Thus uploaded R script (renamed 'txt' for uploading) contains a few small functions that came in handy with that previous context, perhaps useful here.
truncated_normal.txt

@grishaloshadinov @Rick132 @kennedymwavu Do you have any interest in working out a solution to this issue? I am not a user of glmReserve so "I have no horse in this race." If you do not pick up this thread, it will probably remain dormant. But if you do, I will gladly watch and comment.

Dan

[kennedymwavu]

@trinostics Yes, I'd be glad. But I have no knowledge of claims reserving so I'm not sure if I can be of help.

[trinostics]

@kennedymwavu I targeted you in my reply above because, when I saw your comment [Has anyone found workarounds for this issue? I'm facing the same problem in R.], I assumed you had a reserving-related reason to ask. IMO, the two of us without a horse in the race would not likely spearhead a successful solution. Where does your interest originate? More in the glm/bootstrap theory? @grishaloshadinov , as the original poster (OP), are you interested in helping see this through? How about you, @Rick132 [Hello! I have the same problem.] ? If no one is interested in this issue anymore, then it should be closed.

[udaykumarravada]

Sorry for if I miss any information. How can we apply a GLM to negative incremental values i tried for small triangle of 10 X 10 with few negative values I have translated them made the negative values to zero.
But now the triangle is very large which 45 X 45 and has a lot of negative values and what should I do?
Any help would be very appreciated?

[mages]

I think you're trying to force a square peg in a round hole. Consider a different model that is more applicable to your data.

[udaykumarravada]

I think you're trying to force a square peg in a round hole. Consider a different model that is more applicable to your data.

Thanks for the reply.
Can you please suggest some other models ?