Association between two variables over time
Locked
Association between two variables over time
 
|
I would like to investigate the association between serial measurements of
two metric outcomes (e.g., two laboratory parameters measures repeatedly over several years). Depending on the data set, measurements may have been obtained at the same time points for all subjects (study data) or at different time points ('real world' data). Moreover, I may or may not have to perform comparisons between subsets of subjects (defined, e.g., by different treatments received). - What is the most appropriate statistical procedure for this, and - (how) can this be done in SPSS (I use Version 24)? Thank you very much! Andreas -- Sent from: http://spssx-discussion.1045642.n5.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 |
Re: Association between two variables over time
|
|
So, two cases: same timepoints or different time points.
Same timepoints. Timepoints are rows. Column 1 (variable 1) is first outcome. Column 2 (variable 2) is second outcome. Different timepoints. You have to match up outcomes based on their date similarity and then proceed as if the timepoints were the same. It might be quite hard to do this in spss but it depends on the pattern of differences. I think I'd put the data in excel and match up cases according to whether the date difference is less than some value. Good Luck with this! Gene Maguin -----Original Message----- From: SPSSX(r) Discussion <[hidden email]> On Behalf Of AndreasV Sent: Wednesday, June 16, 2021 5:26 AM To: [hidden email] Subject: Association between two variables over time I would like to investigate the association between serial measurements of two metric outcomes (e.g., two laboratory parameters measures repeatedly over several years). Depending on the data set, measurements may have been obtained at the same time points for all subjects (study data) or at different time points ('real world' data). Moreover, I may or may not have to perform comparisons between subsets of subjects (defined, e.g., by different treatments received). - What is the most appropriate statistical procedure for this, and - (how) can this be done in SPSS (I use Version 24)? Thank you very much! Andreas -- Sent from: http://spssx-discussion.1045642.n5.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 |
|
|
In reply to this post by Frank Furter
Tough problem. Maybe one avenue to investigate is to disaggregate the series to a common base. The STATS DISAGG extension command can do this. From its help... CONVERSIONMETHOD (optional) specifies the conversion algorithm. The Chow-Lin, Litterman, and Fernandez methods (with variations) use generalized least squares of the low frequency variable on the aggregated values of the high frequency series and then distribute the differences in the result across the quarterly values. The estimated relationship between the low frequency series (actual and aggregated) is assumed to hold also for the high frequency series. The GLS estimators include estimation of an autoregression parameter, rho. The OLS method does ordinary least squares (with no autocorrelation parameter). These methods all require at least one high frequency indicator. The remaining methods (Denton type) can be used with or without an indicator variable and minimize the sum of squared deviations of the low and high frequency series using two additional parameters. Only one indicator variable is allowed. For details on these algorithms, see http://journal.r-project.org/archive/2013-2/sax-steiner.pdf On Wed, Jun 16, 2021 at 3:26 AM AndreasV <[hidden email]> wrote: I would like to investigate the association between serial measurements of |
|
|
In reply to this post by Frank Furter
If I am reading this correctly, it sounds like Hedeker's Generalized Linear Mixed Models (https://onlinelibrary.wiley.com/doi/abs/10.1002/0470013192.bsa251).
See also: https://hedeker.people.uic.edu/long.html for more on such longitudinal data analysis. I haven't used either, but I thought that SPSS Mixed or GenLinMixed commands were base on Hedeker's analyses. Melissa -----Original Message----- From: SPSSX(r) Discussion <[hidden email]> On Behalf Of AndreasV Sent: Wednesday, June 16, 2021 5:26 AM To: [hidden email] Subject: [SPSSX-L] Association between two variables over time EXTERNAL EMAIL: This email originated from outside of the organization. Do not click any links or open any attachments unless you trust the sender and know the content is safe. I would like to investigate the association between serial measurements of two metric outcomes (e.g., two laboratory parameters measures repeatedly over several years). Depending on the data set, measurements may have been obtained at the same time points for all subjects (study data) or at different time points ('real world' data). Moreover, I may or may not have to perform comparisons between subsets of subjects (defined, e.g., by different treatments received). - What is the most appropriate statistical procedure for this, and - (how) can this be done in SPSS (I use Version 24)? Thank you very much! Andreas -- Sent from: https://gcc02.safelinks.protection.outlook.com/?url=http%3A%2F%2Fspssx-discussion.1045642.n5.nabble.com%2F&data=04%7C01%7CMelissa.Ives%40ct.gov%7Cb8b9082033db4b7f365808d930a8d567%7C118b7cfaa3dd48b9b02631ff69bb738b%7C0%7C0%7C637594323969201455%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&sdata=84aUUvMNwSart5UknR%2B066hQz6GMxJI9L0SzIhE2D6g%3D&reserved=0 ===================== 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 correspondence contains proprietary information some or all of which may be legally privileged; it is for the intended recipient only. If you are not the intended recipient you must not use, disclose, distribute, copy, print, or rely on this correspondence and completely dispose of the correspondence immediately. Please notify the sender if you have received this email in error. NOTE: Messages to or from the State of Connecticut domain may be subject to the Freedom of Information statutes and regulations. ===================== 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 |
Re: Association between two variables over time
|
|
In reply to this post by Frank Furter
Hi, Andreas. I am not sure if I understood your research problem correctly, but the following is my suggestion: For your two variables (say, X and Y), you will have values of X and Y for all your subjects at Time 0 (initial time). This means that you can correlate the X and Y at Time 0. Moreover, you will have a slope of X (say, SLOPE_x) from Time 0 to Time k. You will also have a slope of Y (say, SLOPE_y) from Time 0 to Time k. So, you can test if there is a relationship between SLOPE_x and SLOPE_y. That is, you can test if the change in X is associated with the change in Y. The slopes are latent variables. The two above can be tested using the "Latent Growth Curve Modeling" technique. However, you need the IBM SPSS Amos to carry out the computation. For a video on how to perform the LGCM using Amos, visit this link: http://amosdevelopment.com/features/growth-curve/index.html Just let me know if you have further questions. JOHNNY T. AMORA, MAS Director, Institutional Effectiveness and Research De La Salle-College of Saint Benilde Manila, Philippines Email: [hidden email] || Fb: https://www.facebook.com/amorajohnny1 On Wed, Jun 16, 2021 at 5:26 PM AndreasV <[hidden email]> wrote: I would like to investigate the association between serial measurements of |
|
|
In reply to this post by Frank Furter
Andreas,
This SUGI paper might be helpful: If you decided to use this approach, you’d need to restructure your file in long (vertical) format, fit a multivariate mixed model using the SPSS MIXED procedure, and output the G and R matrices. Using G and R, you would calculate the variance covariance matrix for the observations, typically denoted as V. Recall that: V = Z*G*TRANSPOS(Z) + R where Z = design matrix for the random effects G = covariance matrix for the random effects R = covariance matrix for the residual errors You would then transform the covariance (first off-diagonal element) to a correlation by dividing the covariance by the product of the square root of the variances of the two variables (first two main diagonal elements). HTH. Ryan Sent from my iPhone On Jun 16, 2021, at 5:26 AM, AndreasV <[hidden email]> wrote:
|
«
Return to SPSSX Discussion
|
1 view|%1 views
| Free forum by Nabble | Edit this page |