Question on analysis--many observations per subject

[Allen Frommelt]

I am currently running SPSS 18.  I am trying to determine the best way to use SPSS to analyze our data on remote telemonitoring of blood pressure.  For any individual, I can see a trend though there is a great deal of variability in the data on blood pressure.  I may have up to 900 observations over a period of months on an individual and I have 150 individuals in the pilot study.  The number of observations per individual is too great for the repeated measures ANOVA, but I believe I have to account for the lack of independence in the analysis.  Any ideas would be helpful.  Thanks!

                                                                                        

Healthy Regards,

R. Allen Frommelt

 

This email and all attachments are confidential and intended solely 
for the use of the individual or entity to which they are addressed. 
If you have received this email in error please notify the sender 
by replying to this message. If you are not the intended recipient, 
please delete this message and all attachments immediately.  Do not 
copy, disclose, use or act upon the information contained. Please 
note that any views or opinions presented in this email are solely 
those of the author and do not necessarily represent those of the 
company. Finally, the recipient should check this email and any 
attachments for the presence of viruses. While every attempt is made 
to verify that the contents are safe, the company accepts no liability 
for any damage caused by any virus transmitted by this email.

[luciano basso]

Allen,

Please, Talk more about the purpose of the study? The variability is
an aspect to be considered by itself?

bye bye

Luciano Basso

=====================
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

[Allen Frommelt]

The purpose of the study is to assess the impact that remote telemonitoring has on an individual's ability to help manage blood pressure.  Blood pressure readings are taken multiple times a day and remotely uploaded to our system so that a health coach can reach out to the individual if the blood pressure gets dangerously high or out of control.  What we are seeing is that an individual's blood pressure has tremendous fluctuations throughout the day, and the number of observations per individual is complicating the analysis.

-----Original Message-----
From: luciano basso [mailto:[hidden email]]
Sent: Friday, January 22, 2010 1:40 PM
To: Allen Frommelt
Cc: [hidden email]
Subject: Re: Question on analysis--many observations per subject

Allen,

Please, Talk more about the purpose of the study? The variability is
an aspect to be considered by itself?

bye bye

Luciano Basso

This email and all attachments are confidential and intended solely
for the use of the individual or entity to which they are addressed.
If you have received this email in error please notify the sender
by replying to this message. If you are not the intended recipient,
please delete this message and all attachments immediately.  Do not
copy, disclose, use or act upon the information contained. Please
note that any views or opinions presented in this email are solely
those of the author and do not necessarily represent those of the
company. Finally, the recipient should check this email and any
attachments for the presence of viruses. While every attempt is made
to verify that the contents are safe, the company accepts no liability
for any damage caused by any virus transmitted by this email.

=====================
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

[Steve Simon, P.Mean Consulting]

Allen Frommelt wrote:

> The purpose of the study is to assess the impact that remote
> telemonitoring has on an individual's ability to help manage blood
> pressure.  Blood pressure readings are taken multiple times a day and
> remotely uploaded to our system so that a health coach can reach out
> to the individual if the blood pressure gets dangerously high or out
> of control.  What we are seeing is that an individual's blood
> pressure has tremendous fluctuations throughout the day, and the
> number of observations per individual is complicating the analysis.

There are still a lot of unanswered questions. Is there a concurrent
control group? Do you have measurements prior to the intervention? Are
there any covariates that you need to adjust for?

Still, as a general piece of advice, there are two things you could do.

1. Simplify the model by averaging. Compute a preintervention mean and a
postintervention mean for each subject and then run a paired t-test. Or
compute a postintervention mean for every subject in your treatment
group and a postintervention mean for every subject in your control
group and then use these means as the raw data for a two sample t-test.

2. Learn how to run a mixed model.

Averaging sometimes throws information away, but it also produces a data
set where the statistic of choice is obvious.

Mixed models are messy, but they allow for examination of relationships
and patterns that might otherwise get missed with a simpler model.
--
Steve Simon, Standard Disclaimer
"What do all these numbers mean? Sensitivity,
specificity, and likelihood ratios"
Wednesday, February 17, 11am-noon, CST
Free to all! www.pmean.com/webinars

=====================
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

[Allen Frommelt]

There are still a lot of unanswered questions. Is there a concurrent
control group? [Allen Frommelt: Yes] Do you have measurements prior to the intervention?[Allen Frommelt: Yes] Are
there any covariates that you need to adjust for?[Allen Frommelt: I don't know that they have all been identified yet]

Still, as a general piece of advice, there are two things you could do.

1. Simplify the model by averaging. Compute a preintervention mean and a
postintervention mean for each subject and then run a paired t-test. Or
compute a postintervention mean for every subject in your treatment
group and a postintervention mean for every subject in your control
group and then use these means as the raw data for a two sample t-test.

2. Learn how to run a mixed model.

Averaging sometimes throws information away, but it also produces a data
set where the statistic of choice is obvious.

Mixed models are messy, but they allow for examination of relationships
and patterns that might otherwise get missed with a simpler model.
--
Steve Simon, Standard Disclaimer
"What do all these numbers mean? Sensitivity,
specificity, and likelihood ratios"
Wednesday, February 17, 11am-noon, CST
Free to all! www.pmean.com/webinars

This email and all attachments are confidential and intended solely
for the use of the individual or entity to which they are addressed.
If you have received this email in error please notify the sender
by replying to this message. If you are not the intended recipient,
please delete this message and all attachments immediately.  Do not
copy, disclose, use or act upon the information contained. Please
note that any views or opinions presented in this email are solely
those of the author and do not necessarily represent those of the
company. Finally, the recipient should check this email and any
attachments for the presence of viruses. While every attempt is made
to verify that the contents are safe, the company accepts no liability
for any damage caused by any virus transmitted by this email.

=====================
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

[Maguin, Eugene]

Allan,

Is context for the questions you asked an ongoing, real-time blood pressure
monitoring process where a person's incoming data is 'processed' in some
manner to identify when a coach needs to contact the person for some sort of
intervention work? By 'real-time' I mean that the person's blood pressure
data are uploaded, either manually or by an automated data link, either
immediately or at short, but regular intervals. OR, is this a completed data
collection project in which the data have already been collected and are now
to be analyzed.

If the former, does each person have a predefined blood pressure threshhold
(e.g., 140/100) that trigger a coaching contact? Or, is the threshold a
pattern of blood pressure readings that occur over a day, several days or
weeks?

Gene Maguin



-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Allen Frommelt
Sent: Friday, January 22, 2010 3:22 PM
To: [hidden email]
Subject: Re: Question on analysis--many observations per subject

There are still a lot of unanswered questions. Is there a concurrent
control group? [Allen Frommelt: Yes] Do you have measurements prior to the
intervention?[Allen Frommelt: Yes] Are
there any covariates that you need to adjust for?[Allen Frommelt: I don't
know that they have all been identified yet]

Still, as a general piece of advice, there are two things you could do.

1. Simplify the model by averaging. Compute a preintervention mean and a
postintervention mean for each subject and then run a paired t-test. Or
compute a postintervention mean for every subject in your treatment
group and a postintervention mean for every subject in your control
group and then use these means as the raw data for a two sample t-test.

2. Learn how to run a mixed model.

Averaging sometimes throws information away, but it also produces a data
set where the statistic of choice is obvious.

Mixed models are messy, but they allow for examination of relationships
and patterns that might otherwise get missed with a simpler model.
--
Steve Simon, Standard Disclaimer
"What do all these numbers mean? Sensitivity,
specificity, and likelihood ratios"
Wednesday, February 17, 11am-noon, CST
Free to all! www.pmean.com/webinars

This email and all attachments are confidential and intended solely
for the use of the individual or entity to which they are addressed.
If you have received this email in error please notify the sender
by replying to this message. If you are not the intended recipient,
please delete this message and all attachments immediately.  Do not
copy, disclose, use or act upon the information contained. Please
note that any views or opinions presented in this email are solely
those of the author and do not necessarily represent those of the
company. Finally, the recipient should check this email and any
attachments for the presence of viruses. While every attempt is made
to verify that the contents are safe, the company accepts no liability
for any damage caused by any virus transmitted by this email.

=====================
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

=====================
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

[Allen Frommelt]

The pilot is ongoing, but I'm trying to come up with ideas for the analysis based on how things are starting to be collected and looking at the large number of readings.  The blood pressure data is uploaded through an automated data link.  There is a predefined blood pressure threshold that triggers an alert.

Allen

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Gene Maguin
Sent: Friday, January 22, 2010 3:42 PM
To: [hidden email]
Subject: Re: Question on analysis--many observations per subject

Allan,

Is context for the questions you asked an ongoing, real-time blood pressure
monitoring process where a person's incoming data is 'processed' in some
manner to identify when a coach needs to contact the person for some sort of
intervention work? By 'real-time' I mean that the person's blood pressure
data are uploaded, either manually or by an automated data link, either
immediately or at short, but regular intervals. OR, is this a completed data
collection project in which the data have already been collected and are now
to be analyzed.

If the former, does each person have a predefined blood pressure threshhold
(e.g., 140/100) that trigger a coaching contact? Or, is the threshold a
pattern of blood pressure readings that occur over a day, several days or
weeks?

Gene Maguin



-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Allen Frommelt
Sent: Friday, January 22, 2010 3:22 PM
To: [hidden email]
Subject: Re: Question on analysis--many observations per subject

There are still a lot of unanswered questions. Is there a concurrent
control group? [Allen Frommelt: Yes] Do you have measurements prior to the
intervention?[Allen Frommelt: Yes] Are
there any covariates that you need to adjust for?[Allen Frommelt: I don't
know that they have all been identified yet]

Still, as a general piece of advice, there are two things you could do.

1. Simplify the model by averaging. Compute a preintervention mean and a
postintervention mean for each subject and then run a paired t-test. Or
compute a postintervention mean for every subject in your treatment
group and a postintervention mean for every subject in your control
group and then use these means as the raw data for a two sample t-test.

2. Learn how to run a mixed model.

Averaging sometimes throws information away, but it also produces a data
set where the statistic of choice is obvious.

Mixed models are messy, but they allow for examination of relationships
and patterns that might otherwise get missed with a simpler model.
--
Steve Simon, Standard Disclaimer
"What do all these numbers mean? Sensitivity,
specificity, and likelihood ratios"
Wednesday, February 17, 11am-noon, CST
Free to all! www.pmean.com/webinars

This email and all attachments are confidential and intended solely
for the use of the individual or entity to which they are addressed.
If you have received this email in error please notify the sender
by replying to this message. If you are not the intended recipient,
please delete this message and all attachments immediately.  Do not
copy, disclose, use or act upon the information contained. Please
note that any views or opinions presented in this email are solely
those of the author and do not necessarily represent those of the
company. Finally, the recipient should check this email and any
attachments for the presence of viruses. While every attempt is made
to verify that the contents are safe, the company accepts no liability
for any damage caused by any virus transmitted by this email.

=====================
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

=====================
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

This email and all attachments are confidential and intended solely
for the use of the individual or entity to which they are addressed.
If you have received this email in error please notify the sender
by replying to this message. If you are not the intended recipient,
please delete this message and all attachments immediately.  Do not
copy, disclose, use or act upon the information contained. Please
note that any views or opinions presented in this email are solely
those of the author and do not necessarily represent those of the
company. Finally, the recipient should check this email and any
attachments for the presence of viruses. While every attempt is made
to verify that the contents are safe, the company accepts no liability
for any damage caused by any virus transmitted by this email.

=====================
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

[Maguin, Eugene]

I realize I asked questions that others had asked and you already had
answered. Apologies.

I'm thinking that a person's data could be thought of as consisting of
multiple series of observations in which blood pressure is within limits and
separated by one or more series of observations in which blood pressure
exceeded limits and triggered an intervention. Thought about in this way,
there are several variables that could be constructed but a main one might
be segments consisting of a certain number of observations bracketing the
threshold event. It would seem that a three level model could be set up in
which level 1 models the blood pressure observations. The level 1 model
could be as simple the difference of the pre and post means, or a function
of some covariates and an intervention marker (0,1). I don't know anything
about blood pressure so I have no idea what would be relevant. Level 2 would
be occasion and which refers to the instances of intervention. Level 3 would
be persons.

I'd thought about using a strategy like this for another project but the
project never started. I can point to no examples of an analysis like this
and there may be obvious problems that I don't see.

I'm guessing that you'd like to see that the intervals between intervention
instances increases over time for each person and that the impact of an
intervention either remains the same over time or increases in magnitude
(i.e., a greater and/or quicker blood pressure decrease).




-----Original Message-----
From: Allen Frommelt [mailto:[hidden email]]
Sent: Friday, January 22, 2010 3:49 PM
To: Gene Maguin; [hidden email]
Subject: RE: Question on analysis--many observations per subject

The pilot is ongoing, but I'm trying to come up with ideas for the analysis
based on how things are starting to be collected and looking at the large
number of readings.  The blood pressure data is uploaded through an
automated data link.  There is a predefined blood pressure threshold that
triggers an alert.

Allen

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Gene Maguin
Sent: Friday, January 22, 2010 3:42 PM
To: [hidden email]
Subject: Re: Question on analysis--many observations per subject

Allan,

Is context for the questions you asked an ongoing, real-time blood pressure
monitoring process where a person's incoming data is 'processed' in some
manner to identify when a coach needs to contact the person for some sort of
intervention work? By 'real-time' I mean that the person's blood pressure
data are uploaded, either manually or by an automated data link, either
immediately or at short, but regular intervals. OR, is this a completed data
collection project in which the data have already been collected and are now
to be analyzed.

If the former, does each person have a predefined blood pressure threshhold
(e.g., 140/100) that trigger a coaching contact? Or, is the threshold a
pattern of blood pressure readings that occur over a day, several days or
weeks?

Gene Maguin



-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Allen Frommelt
Sent: Friday, January 22, 2010 3:22 PM
To: [hidden email]
Subject: Re: Question on analysis--many observations per subject

There are still a lot of unanswered questions. Is there a concurrent
control group? [Allen Frommelt: Yes] Do you have measurements prior to the
intervention?[Allen Frommelt: Yes] Are
there any covariates that you need to adjust for?[Allen Frommelt: I don't
know that they have all been identified yet]

Still, as a general piece of advice, there are two things you could do.

1. Simplify the model by averaging. Compute a preintervention mean and a
postintervention mean for each subject and then run a paired t-test. Or
compute a postintervention mean for every subject in your treatment
group and a postintervention mean for every subject in your control
group and then use these means as the raw data for a two sample t-test.

2. Learn how to run a mixed model.

Averaging sometimes throws information away, but it also produces a data
set where the statistic of choice is obvious.

Mixed models are messy, but they allow for examination of relationships
and patterns that might otherwise get missed with a simpler model.
--
Steve Simon, Standard Disclaimer
"What do all these numbers mean? Sensitivity,
specificity, and likelihood ratios"
Wednesday, February 17, 11am-noon, CST
Free to all! www.pmean.com/webinars

This email and all attachments are confidential and intended solely
for the use of the individual or entity to which they are addressed.
If you have received this email in error please notify the sender
by replying to this message. If you are not the intended recipient,
please delete this message and all attachments immediately.  Do not
copy, disclose, use or act upon the information contained. Please
note that any views or opinions presented in this email are solely
those of the author and do not necessarily represent those of the
company. Finally, the recipient should check this email and any
attachments for the presence of viruses. While every attempt is made
to verify that the contents are safe, the company accepts no liability
for any damage caused by any virus transmitted by this email.

=====================
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

=====================
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

This email and all attachments are confidential and intended solely
for the use of the individual or entity to which they are addressed.
If you have received this email in error please notify the sender
by replying to this message. If you are not the intended recipient,
please delete this message and all attachments immediately.  Do not
copy, disclose, use or act upon the information contained. Please
note that any views or opinions presented in this email are solely
those of the author and do not necessarily represent those of the
company. Finally, the recipient should check this email and any
attachments for the presence of viruses. While every attempt is made
to verify that the contents are safe, the company accepts no liability
for any damage caused by any virus transmitted by this email.

=====================
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

[Granaas, Michael]

If the goal of the analysis is patient monitoring it seems like you should be looking at the approaches that businesses use for monitoring manufacturing processes.  They take repeated samples, several times per day/hour and adjust the machinery once some threashold is exceeded.  (That's all I remember, you'll need to talk to someone else for more detail.)

Noisy data might benefit from some sort of data smoothing.  For example a moving average  might provide a stable representation of longer term trends in an individual's blood pressure.

Or am I misunderstanding the goal?

Michael

****************************************************
Michael Granaas             [hidden email]
Assoc. Prof.                Phone: [phone]
Dept. of Psychology         FAX:  [phone]
University of South Dakota
414 E. Clark St.
Vermillion, SD 57069
*****************************************************
________________________________________
From: SPSSX(r) Discussion [[hidden email]] On Behalf Of Allen Frommelt [[hidden email]]
Sent: Friday, January 22, 2010 2:49 PM
To: [hidden email]
Subject: Re: Question on analysis--many observations per subject

The pilot is ongoing, but I'm trying to come up with ideas for the analysis based on how things are starting to be collected and looking at the large number of readings.  The blood pressure data is uploaded through an automated data link.  There is a predefined blood pressure threshold that triggers an alert.

Allen

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Gene Maguin
Sent: Friday, January 22, 2010 3:42 PM
To: [hidden email]
Subject: Re: Question on analysis--many observations per subject

Allan,

Is context for the questions you asked an ongoing, real-time blood pressure
monitoring process where a person's incoming data is 'processed' in some
manner to identify when a coach needs to contact the person for some sort of
intervention work? By 'real-time' I mean that the person's blood pressure
data are uploaded, either manually or by an automated data link, either
immediately or at short, but regular intervals. OR, is this a completed data
collection project in which the data have already been collected and are now
to be analyzed.

If the former, does each person have a predefined blood pressure threshhold
(e.g., 140/100) that trigger a coaching contact? Or, is the threshold a
pattern of blood pressure readings that occur over a day, several days or
weeks?

Gene Maguin



-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Allen Frommelt
Sent: Friday, January 22, 2010 3:22 PM
To: [hidden email]
Subject: Re: Question on analysis--many observations per subject

There are still a lot of unanswered questions. Is there a concurrent
control group? [Allen Frommelt: Yes] Do you have measurements prior to the
intervention?[Allen Frommelt: Yes] Are
there any covariates that you need to adjust for?[Allen Frommelt: I don't
know that they have all been identified yet]

Still, as a general piece of advice, there are two things you could do.

1. Simplify the model by averaging. Compute a preintervention mean and a
postintervention mean for each subject and then run a paired t-test. Or
compute a postintervention mean for every subject in your treatment
group and a postintervention mean for every subject in your control
group and then use these means as the raw data for a two sample t-test.

2. Learn how to run a mixed model.

Averaging sometimes throws information away, but it also produces a data
set where the statistic of choice is obvious.

Mixed models are messy, but they allow for examination of relationships
and patterns that might otherwise get missed with a simpler model.
--
Steve Simon, Standard Disclaimer
"What do all these numbers mean? Sensitivity,
specificity, and likelihood ratios"
Wednesday, February 17, 11am-noon, CST
Free to all! www.pmean.com/webinars

This email and all attachments are confidential and intended solely
for the use of the individual or entity to which they are addressed.
If you have received this email in error please notify the sender
by replying to this message. If you are not the intended recipient,
please delete this message and all attachments immediately.  Do not
copy, disclose, use or act upon the information contained. Please
note that any views or opinions presented in this email are solely
those of the author and do not necessarily represent those of the
company. Finally, the recipient should check this email and any
attachments for the presence of viruses. While every attempt is made
to verify that the contents are safe, the company accepts no liability
for any damage caused by any virus transmitted by this email.

=====================
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

=====================
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

This email and all attachments are confidential and intended solely
for the use of the individual or entity to which they are addressed.
If you have received this email in error please notify the sender
by replying to this message. If you are not the intended recipient,
please delete this message and all attachments immediately.  Do not
copy, disclose, use or act upon the information contained. Please
note that any views or opinions presented in this email are solely
those of the author and do not necessarily represent those of the
company. Finally, the recipient should check this email and any
attachments for the presence of viruses. While every attempt is made
to verify that the contents are safe, the company accepts no liability
for any damage caused by any virus transmitted by this email.

=====================
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

=====================
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

[Kylie]

Hi Allen,

I occasionally work with blood glucose data from continuous glucose
monitoring systems, which provide readings every 5 minutes. Usually we
collect data for 72 hours at a time. Our approach is to define a number of
summary measures depending on the research questions. For example, area
under the curve, time spent above 12 mmol/L, peak value, variability
indices, mean value in the half hour before each meal, mean value in the 4
hours after each meal, etc.

For your study I would recommend considering something similar - identify
what characteristics (eg, mean levels, variability, excursions, number of
times the threshold is reached) of blood pressure you are interested in and
what summaries can be used to measure them. Simplifying the data to a small
number of summary measures will be a big help in tackling this much data.

Cheers,
Kylie.


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Allen Frommelt
Sent: Saturday, 23 January 2010 7:19 AM
To: [hidden email]
Subject: Re: Question on analysis--many observations per subject

The pilot is ongoing, but I'm trying to come up with ideas for the analysis
based on how things are starting to be collected and looking at the large
number of readings.  The blood pressure data is uploaded through an
automated data link.  There is a predefined blood pressure threshold that
triggers an alert.

Allen

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Gene Maguin
Sent: Friday, January 22, 2010 3:42 PM
To: [hidden email]
Subject: Re: Question on analysis--many observations per subject

Allan,

Is context for the questions you asked an ongoing, real-time blood pressure
monitoring process where a person's incoming data is 'processed' in some
manner to identify when a coach needs to contact the person for some sort of
intervention work? By 'real-time' I mean that the person's blood pressure
data are uploaded, either manually or by an automated data link, either
immediately or at short, but regular intervals. OR, is this a completed data
collection project in which the data have already been collected and are now
to be analyzed.

If the former, does each person have a predefined blood pressure threshhold
(e.g., 140/100) that trigger a coaching contact? Or, is the threshold a
pattern of blood pressure readings that occur over a day, several days or
weeks?

Gene Maguin



-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Allen Frommelt
Sent: Friday, January 22, 2010 3:22 PM
To: [hidden email]
Subject: Re: Question on analysis--many observations per subject

There are still a lot of unanswered questions. Is there a concurrent
control group? [Allen Frommelt: Yes] Do you have measurements prior to the
intervention?[Allen Frommelt: Yes] Are
there any covariates that you need to adjust for?[Allen Frommelt: I don't
know that they have all been identified yet]

Still, as a general piece of advice, there are two things you could do.

1. Simplify the model by averaging. Compute a preintervention mean and a
postintervention mean for each subject and then run a paired t-test. Or
compute a postintervention mean for every subject in your treatment
group and a postintervention mean for every subject in your control
group and then use these means as the raw data for a two sample t-test.

2. Learn how to run a mixed model.

Averaging sometimes throws information away, but it also produces a data
set where the statistic of choice is obvious.

Mixed models are messy, but they allow for examination of relationships
and patterns that might otherwise get missed with a simpler model.
--
Steve Simon, Standard Disclaimer
"What do all these numbers mean? Sensitivity,
specificity, and likelihood ratios"
Wednesday, February 17, 11am-noon, CST
Free to all! www.pmean.com/webinars

This email and all attachments are confidential and intended solely
for the use of the individual or entity to which they are addressed.
If you have received this email in error please notify the sender
by replying to this message. If you are not the intended recipient,
please delete this message and all attachments immediately.  Do not
copy, disclose, use or act upon the information contained. Please
note that any views or opinions presented in this email are solely
those of the author and do not necessarily represent those of the
company. Finally, the recipient should check this email and any
attachments for the presence of viruses. While every attempt is made
to verify that the contents are safe, the company accepts no liability
for any damage caused by any virus transmitted by this email.

=====================
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

=====================
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

This email and all attachments are confidential and intended solely
for the use of the individual or entity to which they are addressed.
If you have received this email in error please notify the sender
by replying to this message. If you are not the intended recipient,
please delete this message and all attachments immediately.  Do not
copy, disclose, use or act upon the information contained. Please
note that any views or opinions presented in this email are solely
those of the author and do not necessarily represent those of the
company. Finally, the recipient should check this email and any
attachments for the presence of viruses. While every attempt is made
to verify that the contents are safe, the company accepts no liability
for any damage caused by any virus transmitted by this email.

=====================
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

=====================
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