DynSurv() - Linear prediction for new patients

[pradeepvirdee]

Hi Graeme,

I've fit a joint model using mjoint() and stored the model under the object name "jointmodel". I've obtained dynamic predictions using dynSurv() on the model building sample but also a validation sample containing data from new patients. My aim now is to calculate the model's calibration slope in the validation sample (which involves fitting a Cox model with linear predictions as the only covariate). For the patients used to build the model, I can extract their linear predictions from jointmodel$sfit$linear.predictors. Can I extract the linear prediction for each new patient in the validation sample? If not, is there a way I can work this out?

Pradeep

[graemeleehickey]

I'm not sure I quite follow. jointmodel$sfit is just the Cox model for the separate time-to-event model ignoring the joint-model component, X(t). So jointmodel$sfit$linear.predictors is equivalent to fitting a Cox model with fixed effects and pulling out the linear discriminant term.

…

On Fri, 24 Sept 2021 at 14:10, pradeepvirdee ***@***.***> wrote: Hi Graeme, I've fit a joint model using mjoint() and stored the model under the object name "jointmodel". I've obtained dynamic predictions using dynSurv() on the model building sample but also a validation sample containing data from new patients. My aim now is to calculate the model's calibration slope in the validation sample (which involves fitting a Cox model with linear predictions as the only covariate). For the patients used to build the model, I can extract their linear predictions from jointmodel$sfit$linear.predictors. Can I extract the linear prediction for each new patient in the validation sample? If not, is there a way I can work this out? Pradeep — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#76>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABIG3QLZZMIWUJIQ6AVW65TUDR2CTANCNFSM5EV5XB2Q> .

[pradeepvirdee]

Yes, I figured that out shortly after posting the question. Sorry, at the time I posted, I thought jointmodel$sfit was a Cox model that included the joint model component. Say I wanted linear predictions from the Cox model that does include the joint model component, is that possible? Sorry if this is a silly question. Pradeep Mr Pradeep Virdee NIHR Doctoral Research Fellow; Medical Statistician Centre for Statistics in Medicine (CSM), Botnar Research Centre, NDORMS, Nuffield Orthopaedic Centre, University of Oxford, Windmill Road, Oxford, OX3 7LD Tel: +44 (0)1865 737950 Email: ***@***.***<https://owa.nexus.ox.ac.uk/owa/redir.aspx?C=M9QJvEjCUkuxZIviXHPm2T5wkQ4vR9IIT0wsw9HEKA5FGwCUhzgJWNNdd4XrD5EdWxtXidh_3Ts.&URL=mailto%3apradeep.virdee%40csm.ox.ac.uk> Web: https://www.ndorms.ox.ac.uk/team/pradeep-virdee From: Graeme Hickey ***@***.***> Sent: 16 November 2021 21:49 To: graemeleehickey/joineRML ***@***.***> Cc: Pradeep Virdee ***@***.***>; Author ***@***.***> Subject: Re: [graemeleehickey/joineRML] DynSurv() - Linear prediction for new patients (#76) I'm not sure I quite follow. jointmodel$sfit is just the Cox model for the separate time-to-event model ignoring the joint-model component, X(t). So jointmodel$sfit$linear.predictors is equivalent to fitting a Cox model with fixed effects and pulling out the linear discriminant term.

On Fri, 24 Sept 2021 at 14:10, pradeepvirdee ***@***.***<mailto:***@***.***>> wrote: Hi Graeme, I've fit a joint model using mjoint() and stored the model under the object name "jointmodel". I've obtained dynamic predictions using dynSurv() on the model building sample but also a validation sample containing data from new patients. My aim now is to calculate the model's calibration slope in the validation sample (which involves fitting a Cox model with linear predictions as the only covariate). For the patients used to build the model, I can extract their linear predictions from jointmodel$sfit$linear.predictors. Can I extract the linear prediction for each new patient in the validation sample? If not, is there a way I can work this out? Pradeep — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#76>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABIG3QLZZMIWUJIQ6AVW65TUDR2CTANCNFSM5EV5XB2Q> .

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#76 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ARW3WBHL4G7I3FSZ2LLUFP3UMLGUZANCNFSM5EV5XB2Q>. 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>.

[graemeleehickey]

Since the linear term in the Cox model is a function of a time, you would need to construct it from the coefficients + the time-varying component, m(t).

…

On 17 Nov 2021, at 12:04, pradeepvirdee ***@***.***> wrote: Yes, I figured that out shortly after posting the question. Sorry, at the time I posted, I thought jointmodel$sfit was a Cox model that included the joint model component. Say I wanted linear predictions from the Cox model that does include the joint model component, is that possible? Sorry if this is a silly question. Pradeep Mr Pradeep Virdee NIHR Doctoral Research Fellow; Medical Statistician Centre for Statistics in Medicine (CSM), Botnar Research Centre, NDORMS, Nuffield Orthopaedic Centre, University of Oxford, Windmill Road, Oxford, OX3 7LD Tel: +44 (0)1865 737950 Email: ***@***.***<https://owa.nexus.ox.ac.uk/owa/redir.aspx?C=M9QJvEjCUkuxZIviXHPm2T5wkQ4vR9IIT0wsw9HEKA5FGwCUhzgJWNNdd4XrD5EdWxtXidh_3Ts.&URL=mailto%3apradeep.virdee%40csm.ox.ac.uk> Web: https://www.ndorms.ox.ac.uk/team/pradeep-virdee From: Graeme Hickey ***@***.***> Sent: 16 November 2021 21:49 To: graemeleehickey/joineRML ***@***.***> Cc: Pradeep Virdee ***@***.***>; Author ***@***.***> Subject: Re: [graemeleehickey/joineRML] DynSurv() - Linear prediction for new patients (#76) I'm not sure I quite follow. jointmodel$sfit is just the Cox model for the separate time-to-event model ignoring the joint-model component, X(t). So jointmodel$sfit$linear.predictors is equivalent to fitting a Cox model with fixed effects and pulling out the linear discriminant term. On Fri, 24 Sept 2021 at 14:10, pradeepvirdee ***@***.***<mailto:***@***.***>> wrote: > Hi Graeme, > > I've fit a joint model using mjoint() and stored the model under the > object name "jointmodel". I've obtained dynamic predictions using dynSurv() > on the model building sample but also a validation sample containing data > from new patients. My aim now is to calculate the model's calibration slope > in the validation sample (which involves fitting a Cox model with linear > predictions as the only covariate). For the patients used to build the > model, I can extract their linear predictions from > jointmodel$sfit$linear.predictors. Can I extract the linear prediction for > each new patient in the validation sample? If not, is there a way I can > work this out? > > Pradeep > > — > You are receiving this because you are subscribed to this thread. > Reply to this email directly, view it on GitHub > <#76>, or unsubscribe > <https://github.com/notifications/unsubscribe-auth/ABIG3QLZZMIWUJIQ6AVW65TUDR2CTANCNFSM5EV5XB2Q> > . > — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#76 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ARW3WBHL4G7I3FSZ2LLUFP3UMLGUZANCNFSM5EV5XB2Q>. 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 commented. Reply to this email directly, view it on GitHub <#76 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABIG3QOKO4ORJAU6UIVKS63UMOK3RANCNFSM5EV5XB2Q>.