confidence intervals

[drfg2008]

We use a linear regression model as a time series model on a daily basis. For each day the predicted value PRE and the LMCI, UMCI are saved (/SAVE PRED MCIN). The confidence level is 95%.
How can we estimate the LMCI and UMCI on a monthly basis: Would it be correct to compute the sum of all LMCI of a month (on a daily basis) to generate the monthly LMCI and the sum of all UMCI on a daily basis of a month to generate the monthly UMCI?
Thanks

[salil deo]

While using the linear regression model I am sure you would be having a sample size <N> for each day ; I my opinion simply averaging can be done but would not be absolutely accurate ; more perfect number would be obtained by obtaining a weighted mean ... this can be done easily in R ... the formula would be :

if each sample is n1,n2.... and each CI is CI1,CI2... then total sample size = n1+n2.... = N

weighted mean = (n1/N)*CI1 + (n2/N)*CI2 + ... 

it can be easily prepared and calculated using Excel spreadsheet

hope this helps .... any opinions ??? 

 

Dr Salil V Deo

Consultant Cardiovascular Surgeon

Fellow, Cardiac Transplantation, Mayo Clinic, USA (2012)

Adventist Wockhardt Heart Hospital

Athawalines, Surat 395001

---

From: drfg2008 <[hidden email]>
To: [hidden email]
Sent: Sunday, 4 August 2013 12:21 PM
Subject: confidence intervals

We use a linear regression model as a time series model on a daily basis.
For each day the predicted value PRE and the LMCI, UMCI are saved (/SAVE
PRED MCIN). The confidence level is 95%.
How can we estimate the LMCI and UMCI on a monthly basis: Would it be
correct to compute the sum of all LMCI of a month (on a daily basis) to
generate the monthly LMCI and the sum of all UMCI on a daily basis of a
month to generate the monthly UMCI?
Thanks




-----
Dr. Frank Gaeth
FU-Berlin

--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/confidence-intervals-tp5721474.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to
LISTSERV@... (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Jon K Peck]

I don't think this is a good way to approach this problem.  The prediction intervals are likely to be highly correlated, and neither of these approaches takes that into consideration.

The obvious way to do this would be to aggregate the data and estimate a monthly equation.  In fact, the monthly model might well be different, since daily effects would be washed out.  If that is operationally impractical, you might at least experiment with different approximations by estimating a monthly model and comparing those CIs with the various approximations.


Jon Peck (no "h") aka Kim
Senior Software Engineer, IBM
[hidden email]
phone: [phone]




From:        salil deo <[hidden email]>
To:        [hidden email],
Date:        08/04/2013 01:34 AM
Subject:        Re: [SPSSX-L] confidence intervals
Sent by:        "SPSSX(r) Discussion" <[hidden email]>

---

While using the linear regression model I am sure you would be having a sample size <N> for each day ; I my opinion simply averaging can be done but would not be absolutely accurate ; more perfect number would be obtained by obtaining a weighted mean ... this can be done easily in R ... the formula would be :

if each sample is n1,n2.... and each CI is CI1,CI2... then total sample size = n1+n2.... = N

weighted mean = (n1/N)*CI1 + (n2/N)*CI2 + ...

it can be easily prepared and calculated using Excel spreadsheet

hope this helps .... any opinions ???
 
Dr Salil V Deo
Consultant Cardiovascular Surgeon
Fellow, Cardiac Transplantation, Mayo Clinic, USA (2012)
Adventist Wockhardt Heart Hospital
Athawalines, Surat 395001

---

From: drfg2008 <[hidden email]>
To: [hidden email]
Sent: Sunday, 4 August 2013 12:21 PM
Subject: confidence intervals

We use a linear regression model as a time series model on a daily basis.
For each day the predicted value PRE and the LMCI, UMCI are saved (/SAVE
PRED MCIN). The confidence level is 95%.
How can we estimate the LMCI and UMCI on a monthly basis: Would it be
correct to compute the sum of all LMCI of a month (on a daily basis) to
generate the monthly LMCI and the sum of all UMCI on a daily basis of a
month to generate the monthly UMCI?
Thanks




-----
Dr. Frank Gaeth
FU-Berlin

--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/confidence-intervals-tp5721474.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[drfg2008]

Thanks a lot! Just one more information: Since the model is 'event-driven' and the events are not correlated the values (and especially the residuals of the model) are also almost not (auto-) correlated. A monthly model is not possible, because of too many prediction variables in the model.

[Maurice Vergeer]

then the question is why you would want the monthly confidence
intervals in the first place?


On Sun, Aug 4, 2013 at 7:14 PM, drfg2008 <[hidden email]> wrote:

> Thanks a lot! Just one more information: Since the model is 'event-driven'
> and the events are not correlated the values (and especially the residuals
> of the model) are also almost not (auto-) correlated. A monthly model is not
> possible, because of too many prediction variables in the model.
>
>
>
>
>
>
>
> -----
> Dr. Frank Gaeth
> FU-Berlin
>
> --
> View this message in context: http://spssx-discussion.1045642.n5.nabble.com/confidence-intervals-tp5721474p5721480.html
> Sent from the SPSSX Discussion mailing list archive at Nabble.com.
>
> =====================
> To manage your subscription to SPSSX-L, send a message to
> [hidden email] (not to SPSSX-L), with no body text except the
> command. To leave the list, send the command
> SIGNOFF SPSSX-L
> For a list of commands to manage subscriptions, send the command
> INFO REFCARD

--
________________________________________________
Maurice Vergeer
To contact me, see http://mauricevergeer.nl/node/5
To see my publications, see http://mauricevergeer.nl/node/1
________________________________________________

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD