SPSS Syntax - Linear Interpolation

[JMA1234]

CONTENTS DELETED

The author has deleted this message.

[JMA1234]

CONTENTS DELETED

The author has deleted this message.

[Bruce Weaver]

Please provide a small example of what the data file looks like now, and what you want it to look like after.  

AK1234 wrote

I have 7 data points (time vs. melatonin concentration) for each ID number (in long format). I need to linear interpolate the time at two different set melatonin concentrations and then loop this through all ID numbers. If anyone has any suggestions on how I can do this, that would be greatly appreciated.

[JMA1234]

CONTENTS DELETED

The author has deleted this message.

[Mike]

I have to second Bruce's request because it hard to understand what

your data actually is.  You say you have long format data which

means you have 7 cases per ID but "time vs melatonin concentration"

make no sense.  Consider the following data layout for one ID:

 

ID	time	mel_conc
001	1	202
001	2	222
001	3	241
001	4	269
001	5	280
001	6	299
001	7	320

 

Now this is all I can infer and I don't have much confidence that it

is completely correct (sorry, my ESP doesn't seem to be working today).

I assume that there is a positive linear relationship between mel_conc

and time which suggests that you can get a simple regression equation

to predict the value of mel_conc for any value of time (it is unclear

when you have actual times or ordered time points like above which

might be interpreted as the original time converted into ranks).

 

But I have to ask, why do you want to do linear interpolation?

 

My first inclination is that you have a melatonin concentration

for an ID but not the time, as follows:

 

ID	time	mel_conc
001	3	241
001		269
001	5	280

 

Mel_conc=269 falls between the values of 241 and 280 and you

might want the time that corresponds to the missing time (if the

time is original time units and not ranks).  But you don't say

anything about missing data which implies you might be interested

in some time value between the times you (say, 3.5).  Again,

a simple regression equation would allow you to get this value

(run regression using split files using ID as the splitting variable)

but I still don't understand whether this is what you wnat to do

or why you might want to do it.

 

-Mike Palij

New York University

[hidden email]

 

 

On Saturday, May 27, 2017 1:28 PM, Bruce Weaver wrote:

> Please provide a small example of what the data file looks like now, and what
> you want it to look like after.  
>
> AK1234 wrote
>> I have 7 data points (time vs. melatonin concentration) for each ID number
>> (in long format). I need to linear interpolate the time at two different
>> set melatonin concentrations and then loop this through all ID numbers. If
>> anyone has any suggestions on how I can do this, that would be greatly
>> appreciated.

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

[Bruce Weaver]

I don't see a variable called concentration.  Do you mean you want to compute the mean of the time values for samples 3 and 4 for each subject?  

By the way, Nabble shows that your posts have "NOT been accepted by the mailing list yet."  (Because of that, I suspect Mike had not seen your reply when he made his post.)  Have you joined the actual mailing list?  

AK1234 wrote

This is how the data looks currently for two participants. Each participant has data for 7 samples in long format. I would like to linear interpolate the time at a melatonin concentration of 3 and a melatonin concentration of 4 for each participant. I have a very large N so it would also be necessary to use some sort of looping function.

[Mike]

Note: the OP's posts are not making it to the mailing list

and I just checked Nabble to see the thread and I saw

that the OP has responded and uses a data layout much

like the one I use below.

 

After looking at you data table I think the question you

are asking is "what are the predicted times for melatonin

concentrations of 3 (units) and 4 (units)" which suggests

that you want to regression time in original units on

melatonin concentration [time = b_0 + b_1(mel_conc)].

Again, I think you may be able to this by using split files

and calculating the simple regression for each ID which

you can use to predict/calculate the time for mel_conc = 3 and 4.

The split files makes getting the regression equation simple

but then using them to calculate the melatonin concentration

is going to either require additional programming or hand

calculation.

 

A couple of notes:

(1)  People who are more familiar with multilevel analysis

can answer whether such an analysis would be appropriate

here in order to determine the regression of time on

melatonin is the same for all IDs or different across IDs.

If the regressions do not vary significant, then a single

equation could be used for all IDs.  If there is significant

difference among the equations (e.g. different slopes),

then you'll have to calculate the time on a ID level.

 

(2) A quick look at your data raises the question whether

melatonin actually has a linear relationship with time (or

vice versa).  Some IDs may have a quadratic relationship

or more complex pattern (theoretically, I believe, up to

6th order polynomial).

 

I'd like to see other people's suggestions.

 

-Mike Palij

New York University

[hidden email]

 

 

 

----- Original Message -----

From: [hidden email]

To: [hidden email] ; [hidden email]

Cc: [hidden email]

Sent: Saturday, May 27, 2017 2:02 PM

Subject: Re: SPSS Syntax - Linear Interpolation

I have to second Bruce's request because it hard to understand what

your data actually is.  You say you have long format data which

means you have 7 cases per ID but "time vs melatonin concentration"

make no sense.  Consider the following data layout for one ID:

 

ID	time	mel_conc
001	1	202
001	2	222
001	3	241
001	4	269
001	5	280
001	6	299
001	7	320

 

Now this is all I can infer and I don't have much confidence that it

is completely correct (sorry, my ESP doesn't seem to be working today).

I assume that there is a positive linear relationship between mel_conc

and time which suggests that you can get a simple regression equation

to predict the value of mel_conc for any value of time (it is unclear

when you have actual times or ordered time points like above which

might be interpreted as the original time converted into ranks).

 

But I have to ask, why do you want to do linear interpolation?

 

My first inclination is that you have a melatonin concentration

for an ID but not the time, as follows:

 

ID	time	mel_conc
001	3	241
001		269
001	5	280

 

Mel_conc=269 falls between the values of 241 and 280 and you

might want the time that corresponds to the missing time (if the

time is original time units and not ranks).  But you don't say

anything about missing data which implies you might be interested

in some time value between the times you (say, 3.5).  Again,

a simple regression equation would allow you to get this value

(run regression using split files using ID as the splitting variable)

but I still don't understand whether this is what you wnat to do

or why you might want to do it.

 

-Mike Palij

New York University

[hidden email]

 

 

On Saturday, May 27, 2017 1:28 PM, Bruce Weaver wrote:

> Please provide a small example of what the data file looks like now, and what
> you want it to look like after.  
>
> AK1234 wrote
>> I have 7 data points (time vs. melatonin concentration) for each ID number
>> (in long format). I need to linear interpolate the time at two different
>> set melatonin concentrations and then loop this through all ID numbers. If
>> anyone has any suggestions on how I can do this, that would be greatly
>> appreciated.

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

[JMA1234]

CONTENTS DELETED

The author has deleted this message.

[Maguin, Eugene]

In looking at the two example cases, the relationship doesn't look too linear. That said, a linear assumption may be preferred. If so, that preference leads down one syntax path. Another syntax path is to interpolate the between points enclosing an interval that includes 3 or includes 4. These two different paths probably, almost certainly will give different time codes. Which way do you want to go?
Gene Maguin


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of AK1234
Sent: Saturday, May 27, 2017 2:25 PM
To: [hidden email]
Subject: Re: SPSS Syntax - Linear Interpolation

Hello Bruce,

I should have clarified. MelaAvg corresponds with the melatonin concentration. I would like to linear interpolate the time at which a melatonin concentration of 3 and a melatonin of concentration of 4 are hit/surpassed.

I have received confirmation that I was subscribed to the mailing list. I am unsure as to why it continues to say that my posts have not be accepted. I will look into this. Thank you for pointing that out.



--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/SPSS-Syntax-Linear-Interpolation-tp5734282p5734290.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

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

[JMA1234]

CONTENTS DELETED

The author has deleted this message.

[Art Kendall]

CASESTOVARS is most likely part of the process.

After doing CASESTOVARS for purposes of visualization before doing any analysis, I suggest you plot a line per person with concentration on the Y axis and "time" on the X axis.  You may want to break the data into sets of cases for the plots. Do the shapes appear at least vaguely similar? How do loess fits look?

Is there reason to believe the shapes should be similar?

Is there reason to believe the shape is linear?

Are there subgroups of cases that you wish to compare/contrast their fit?

For your two example cases, what value do you want the syntax to create?

[Bruce Weaver]

JMA, no matter how urgent this may be for you, asking the same question repeatedly is not going to make anyone respond more quickly!  Bear in mind that it is the weekend, and a long weekend (Memorial Day) for our American friends.  So some potential respondents may not even see your question until Tuesday.  

Note too that using different subject lines when asking the same question clutters up the archive unnecessarily.  

JMA1234 wrote

I am having trouble figuring out how to derive an equation of a line using SPSS syntax. More specifically, my data is in long format, with 7 data points (time vs. melatonin concentration) for each ID number. I would like to derive an equation for each individual based on their data points and then be able to compute predicted values for new points based on this equation. I hope that my question makes sense. Any assistance in this matter would be greatly appreciated.

[JMA1234]

CONTENTS DELETED

The author has deleted this message.

[Rich Ulrich]

I think I would be confused a bit, even if I had read all the posts in sequence from Nabble.

So, I confess that I'm not sure of the ultimate end.

But if I wanted the regression for each person with 7 (always, exactly) points, 1 to 7,

I would have all the 7 lines of data in one record and construct the simple linear contrast

by hand, for (-3 to 3).   The covariance for the contrast term can be written like

      3*(Mtime7- Mtime1) + 2*(Mtime6-Mtime2) + (Mtime5-Mtime3)

and the sum of squares of the contrasts is 28.  Compute the mean and sum of squares of

what I've labeled Mtime_   and you have all the values for the regression coefficient and the

intercept. 

--

Rich Ulrich

---

From: SPSSX(r) Discussion <[hidden email]> on behalf of JMA1234 <[hidden email]>
Sent: Sunday, May 28, 2017 9:22:53 AM
To: [hidden email]
Subject: Re: SPSS Syntax - Equation of Line

 

I would like to produce an equation for a line for each persons data based on
their points. Essentially I would like to create a variable indicating the
participant's slope of their line as well as another variable indicating the
y-intercept of that line.



--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/SPSS-Syntax-Linear-Interpolation-tp5734282p5734297.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

===================== 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]

AK,

One method, the one that you indicated is customary practice, works like this.

I’ll presume the file is appropriately sorted. Then, casestovars by id to get a wide format file (drop sample).

Per the example data you supplied.

Id mel.1 mel.2 mel.3 mel.4 mel.5 mel.6 mel.7 time.1 … time.7

1 2.0 2.5 3.5 7.0 9.0 10.0 17.0 23:30 …. 26:30

2 1.0 3.0 7.0 8.0 13.0 17.0 30.0 18:00 … 21:00

 

Vector mel=mel.1 to mel.7/time=time.1 to time.7.

Loop #i=1 to 6.

*look for the lower and upper bounds on the interval around 3.0.

+  do if (mel(7) lt 3.0).

+     break.

+  else if (mel(#i) eq 3.0).

+     compute m3.0time=time(#i).

+  else if (mel(#i+1) eq 3.0).

+     compute m3.0time=time(#i+1).

+  else if (mel(#i) lt 3.0 and mel(#i+1) gt 3).

+     compute m3.0time=time(#i)+((time(#i+1)-time(#i))*((3.0-mel(#i))/

      (mel(#i+1)-mel(#i)))).

+  end if.

end loop.

format m3.0time(time5).

execute.

 

You can modify the code to look for the mel=4.0 time point.

 

You’d also mentioned looking for the within person slope and intercept. I agree with Mike’s  advice to use a split files command. Depending how many cases (not records) you have, you may find it useful to use the OMS-OMSEND commands.

 

Gene Maguin

 

 

 

 

 

 

 

 

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Rich Ulrich
Sent: Sunday, May 28, 2017 2:02 PM
To: [hidden email]
Subject: Re: SPSS Syntax - Equation of Line

 

I think I would be confused a bit, even if I had read all the posts in sequence from Nabble.

So, I confess that I'm not sure of the ultimate end.

 

But if I wanted the regression for each person with 7 (always, exactly) points, 1 to 7,

I would have all the 7 lines of data in one record and construct the simple linear contrast

by hand, for (-3 to 3).   The covariance for the contrast term can be written like

 

      3*(Mtime7- Mtime1) + 2*(Mtime6-Mtime2) + (Mtime5-Mtime3)

and the sum of squares of the contrasts is 28.  Compute the mean and sum of squares of

what I've labeled Mtime_   and you have all the values for the regression coefficient and the

intercept. 

 

--

Rich Ulrich

 

---

From: SPSSX(r) Discussion <[hidden email]> on behalf of JMA1234 <[hidden email]>
Sent: Sunday, May 28, 2017 9:22:53 AM
To: [hidden email]
Subject: Re: SPSS Syntax - Equation of Line

 

I would like to produce an equation for a line for each persons data based on
their points. Essentially I would like to create a variable indicating the
participant's slope of their line as well as another variable indicating the
y-intercept of that line.



--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/SPSS-Syntax-Linear-Interpolation-tp5734282p5734297.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

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

[David Marso]

Dear OP (JMA1234).
Please STOP deleting the contents of your questions.
This entire thread has bcome chaotic mess due to what we would term unacceptable netiquette .
Get with the program or leave.
As far as I can tell you merely need to do SPIT FILE BY ID and run a bloody regression!