GEE: Binary logistic with repeated measures
Locked
 
|
I'm trying to run a GEE analysis in SPSS 16 and not getting very far at this point as this is new territory for me. I am trying to analyze the probability of pursuit upon encounter for individual hunters in the Amazon of Bolivia. The repeated measure is the ID of each hunter. Each hunter was followed anywhere from one to five times while in the forest (trip id is repeated to every time there is a new encounter). On each of his follows he may or may not have encountered any number of animals which he then must decide whether or not to pursue. As it stands now I am trying to use the ID as the subject variable but keep getting the error message:
Warnings There are at least two records with the same values for the subject and within-subject variables. No output will be displayed. This command is not executed. How does my data need to be structured in order to run this type of analysis given the nature of my data? Any insight would be greatly appreciated. Cheers, Allen Gillespie University of Auckland |
Re: GEE: Binary logistic with repeated measures
|
|
Have a look at the wheeze_steubenville.sav dataset that ships with SPSS and the associated Case Study (Help > Case Studies... and navigate to the GEE section, or click on the "Show Me" link in the main GEE help topic). That should show you the data structure for a repeated measures binary logistic.
Alex -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of alleng Sent: Wednesday, March 05, 2008 7:41 PM To: [hidden email] Subject: GEE: Binary logistic with repeated measures I'm trying to run a GEE analysis in SPSS 16 and not getting very far at this point as this is new territory for me. I am trying to analyze the probability of pursuit upon encounter for individual hunters in the Amazon of Bolivia. The repeated measure is the ID of each hunter. Each hunter was followed anywhere from one to five times while in the forest (trip id is repeated to every time there is a new encounter). On each of his follows he may or may not have encountered any number of animals which he then must decide whether or not to pursue. As it stands now I am trying to use the ID as the subject variable but keep getting the error message: Warnings There are at least two records with the same values for the subject and within-subject variables. No output will be displayed. This command is not executed. How does my data need to be structured in order to run this type of analysis given the nature of my data? Any insight would be greatly appreciated. Cheers, Allen Gillespie University of Auckland -- View this message in context: http://www.nabble.com/GEE%3A-Binary-logistic-with-repeated-measures-tp15864680p15864680.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 |
|
|
Thanks Alex, that was helpful. Unfortunately, the program still won't let me analyze the data set because it is telling me that there are at least two records with the same values for the subject and within-subject variables. I'm not sure if this is a data error on my part or due to the structure of my data. In the wheeze example each subject has four measures, however, in my data the number of encounters (cases) is highly variable so that one individual may be represented 12 times and another may be represented 45 times depending not only on how many times they were followed in the field but also how many times they actually were successful in encountering an animal. Could this be the reason that it is not permitting the analysis?
Cheers, Allen
|
Re: GEE: Binary logistic with repeated measures
|
|
Allen,
I have no experience with the spss GEE module. However, can you rule out the obvious? That is, that there are two records with the same values for the subject and within-subject variables. If you have some later verison of spss, there is an identify duplicates function on the data menu dropdown. If you don't have that, the checking is harder but not impossible. Repost and I or others can help you with syntax. Gene Maguin ===================== 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 |
|
|
Hi Gene,
Yes, I've been trying to diagnose this problem all afternoon. The subject numeral is a five to six digit number and the within subject variable is a two digit numeral. I've even gone as far as dissecting the data set to simplify it and I still get the same warning message. I initially thought it might be a problem with extra cases and missing values but can safely rule that out. I can run an analysis using a different dataset (the above example, wheeze) so it has to be something with the nature of my data. I've also tried creating a new ID variable again to no avail. Cheers, Allen
|
Re: GEE: Binary logistic with repeated measures
|
|
In reply to this post by alleng-2
Hmmm. Let me make sure I understand your data setup. You have variables like:
Hunter Trip Encounter Pursuit 1 1 1 No 1 1 2 Yes 1 1 3 No 1 2 1 No 1 2 2 No 1 3 1 No 1 3 2 No 1 3 3 No 1 3 4 Yes 2 1 1 No .... And so on, where Hunter ranges from 1 to the number of hunters; within Hunters, Trip ranges from 1 to the number of times each Hunter was followed; within Trips, Encounter ranges from 1 to the number animals encountered on each Trip. Pursuit marks whether or not the animal was pursued. You may have other factors and covariates that further describe the Hunter, Trip, or Encounter. Does your REPEATED subcommand then look something like: /REPEATED SUBJECT=Hunter WITHINSUBJECT=Trip*Encounter ? If so, and you're getting the warning, then there is a duplicate combination of Hunter*Trip*Encounter that needs to be corrected. If not, then what is your actual data setup and REPEATED subcommand? Also, if I've understood correctly, even once you have this running properly GENLIN will be estimating a 45x45 working correlation matrix, which will likely cause its own problems. But get this to run before worrying about that. Alex -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of alleng Sent: Thursday, March 06, 2008 2:54 PM To: [hidden email] Subject: Re: GEE: Binary logistic with repeated measures Thanks Alex, that was helpful. Unfortunately, the program still won't let me analyze the data set because it is telling me that there are at least two records with the same values for the subject and within-subject variables. I'm not sure if this is a data error on my part or due to the structure of my data. In the wheeze example each subject has four measures, however, in my data the number of encounters (cases) is highly variable so that one individual may be represented 12 times and another may be represented 45 times depending not only on how many times they were followed in the field but also how many times they actually were successful in encountering an animal. Could this be the reason that it is not permitting the analysis? Cheers, Allen Reutter, Alex wrote: > > Have a look at the wheeze_steubenville.sav dataset that ships with SPSS > and the associated Case Study (Help > Case Studies... and navigate to the > GEE section, or click on the "Show Me" link in the main GEE help topic). > That should show you the data structure for a repeated measures binary > logistic. > > Alex > > > -----Original Message----- > From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of > alleng > Sent: Wednesday, March 05, 2008 7:41 PM > To: [hidden email] > Subject: GEE: Binary logistic with repeated measures > > I'm trying to run a GEE analysis in SPSS 16 and not getting very far at > this > point as this is new territory for me. I am trying to analyze the > probability of pursuit upon encounter for individual hunters in the Amazon > of Bolivia. The repeated measure is the ID of each hunter. Each hunter > was > followed anywhere from one to five times while in the forest (trip id is > repeated to every time there is a new encounter). On each of his follows > he > may or may not have encountered any number of animals which he then must > decide whether or not to pursue. As it stands now I am trying to use the > ID > as the subject variable but keep getting the error message: > > Warnings > There are at least two records with the same values for the subject and > within-subject variables. No output will be displayed. > This command is not executed. > > How does my data need to be structured in order to run this type of > analysis > given the nature of my data? > > Any insight would be greatly appreciated. > > Cheers, > Allen Gillespie > University of Auckland > > -- > View this message in context: > http://www.nabble.com/GEE%3A-Binary-logistic-with-repeated-measures-tp15864680p15864680.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 |
|
|
Yes, essentially except that the hunter ID is a five or six digit number that is non-sequential and the trip ID is also non-sequential (1 through 150) in that the hunter may have been followed on the 14th, 47th, and 65th coded trips. Pursue is dichotomous 1, 0.
Cheers, Allen
|
Re: GEE: Binary logistic with repeated measures
|
|
Okay, then can you run Data > Identify Duplicate Cases..., select Hunter, Trip, and Encounter as the variable to "Define Matching Cases by", and click OK? If this doesn't turn up the cases causing the warning, then I'm stumped.
Alex -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of alleng Sent: Saturday, March 08, 2008 8:29 PM To: [hidden email] Subject: Re: GEE: Binary logistic with repeated measures Yes, essentially except that the hunter ID is a five or six digit number that is non-sequential and the trip ID is also non-sequential (1 through 150) in that the hunter may have been followed on the 14th, 47th, and 65th coded trips. Pursue is dichotomous 1, 0. Cheers, Allen Reutter, Alex wrote: > > Hmmm. Let me make sure I understand your data setup. You have variables > like: > > Hunter Trip Encounter Pursuit > 1 1 1 No > 1 1 2 Yes > 1 1 3 No > 1 2 1 No > 1 2 2 No > 1 3 1 No > 1 3 2 No > 1 3 3 No > 1 3 4 Yes > 2 1 1 No > .... > > And so on, where Hunter ranges from 1 to the number of hunters; within > Hunters, Trip ranges from 1 to the number of times each Hunter was > followed; within Trips, Encounter ranges from 1 to the number animals > encountered on each Trip. Pursuit marks whether or not the animal was > pursued. You may have other factors and covariates that further describe > the Hunter, Trip, or Encounter. > > Does your REPEATED subcommand then look something like: > > /REPEATED SUBJECT=Hunter WITHINSUBJECT=Trip*Encounter > > ? > > If so, and you're getting the warning, then there is a duplicate > combination of Hunter*Trip*Encounter that needs to be corrected. > > If not, then what is your actual data setup and REPEATED subcommand? > > Also, if I've understood correctly, even once you have this running > properly GENLIN will be estimating a 45x45 working correlation matrix, > which will likely cause its own problems. But get this to run before > worrying about that. > > Alex > > > -----Original Message----- > From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of > alleng > Sent: Thursday, March 06, 2008 2:54 PM > To: [hidden email] > Subject: Re: GEE: Binary logistic with repeated measures > > Thanks Alex, that was helpful. Unfortunately, the program still won't let > me > analyze the data set because it is telling me that there are at least two > records with the same values for the subject and within-subject variables. > I'm not sure if this is a data error on my part or due to the structure of > my data. In the wheeze example each subject has four measures, however, > in > my data the number of encounters (cases) is highly variable so that one > individual may be represented 12 times and another may be represented 45 > times depending not only on how many times they were followed in the field > but also how many times they actually were successful in encountering an > animal. Could this be the reason that it is not permitting the analysis? > > Cheers, > Allen > > > > > Reutter, Alex wrote: >> >> Have a look at the wheeze_steubenville.sav dataset that ships with SPSS >> and the associated Case Study (Help > Case Studies... and navigate to the >> GEE section, or click on the "Show Me" link in the main GEE help topic). >> That should show you the data structure for a repeated measures binary >> logistic. >> >> Alex >> >> >> -----Original Message----- >> From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of >> alleng >> Sent: Wednesday, March 05, 2008 7:41 PM >> To: [hidden email] >> Subject: GEE: Binary logistic with repeated measures >> >> I'm trying to run a GEE analysis in SPSS 16 and not getting very far at >> this >> point as this is new territory for me. I am trying to analyze the >> probability of pursuit upon encounter for individual hunters in the >> Amazon >> of Bolivia. The repeated measure is the ID of each hunter. Each hunter >> was >> followed anywhere from one to five times while in the forest (trip id is >> repeated to every time there is a new encounter). On each of his follows >> he >> may or may not have encountered any number of animals which he then must >> decide whether or not to pursue. As it stands now I am trying to use the >> ID >> as the subject variable but keep getting the error message: >> >> Warnings >> There are at least two records with the same values for the subject and >> within-subject variables. No output will be displayed. >> This command is not executed. >> >> How does my data need to be structured in order to run this type of >> analysis >> given the nature of my data? >> >> Any insight would be greatly appreciated. >> >> Cheers, >> Allen Gillespie >> University of Auckland >> >> -- >> View this message in context: >> http://www.nabble.com/GEE%3A-Binary-logistic-with-repeated-measures-tp15864680p15864680.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 > > -- View this message in context: http://www.nabble.com/GEE%3A-Binary-logistic-with-repeated-measures-tp15864680p15923209.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 |
|
|
Hi Alex,
It turns up one duplicate so I guess it's time to put my nose to the screen and see if I can't find it. Cheers for all input. Allen
|
Re: GEE: Binary logistic with repeated measures
|
At 12:27 AM 3/11/2008, alleng wrote:
>It turns up one duplicate so I guess it's time to put my nose to the >screen and see if I can't find it. Cheers for all input. No nose to the screen! To find a duplicate in a single variable VAR, AGGREGATE OUTFILE=* MODE = ADDVARIABLES /BREAK=VAR /N_INST 'No. times this value of VAR occurs' = NU. TEMPORARY. SELECT IF N_INST GT 1. LIST. -- No virus found in this outgoing message. Checked by AVG. Version: 7.5.518 / Virus Database: 269.21.7/1324 - Release Date: 3/10/2008 7:27 PM ===================== 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 |
|
|
Since you used Identify Duplicate Cases to find the duplicates, the default behavior is to move them to to the top of file. Just look in the Data Editor.
IDC will also create a matchsequence variable that counts within duplicates, so you can look for a number bigger than one if you checked that box. -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Richard Ristow Sent: Tuesday, March 11, 2008 9:53 AM To: [hidden email] Subject: Re: [SPSSX-L] GEE: Binary logistic with repeated measures At 12:27 AM 3/11/2008, alleng wrote: >It turns up one duplicate so I guess it's time to put my nose to the >screen and see if I can't find it. Cheers for all input. No nose to the screen! To find a duplicate in a single variable VAR, AGGREGATE OUTFILE=* MODE = ADDVARIABLES /BREAK=VAR /N_INST 'No. times this value of VAR occurs' = NU. TEMPORARY. SELECT IF N_INST GT 1. LIST. -- No virus found in this outgoing message. Checked by AVG. Version: 7.5.518 / Virus Database: 269.21.7/1324 - Release Date: 3/10/2008 7:27 PM ===================== 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 |
|
|
Yes, sure enough. I ran and walked away yesterday evening (sometimes a good idea) only to arrive and find my duplicates staring me in the face upon arrival this morning. Many thanks to everyone.
Cheers, Allen
|
«
Return to SPSSX Discussion
|
1 view|%1 views
| Free forum by Nabble | Edit this page |