TREE syntax that works doesn't create a tree for a subset (n=2557)

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

TREE syntax that works doesn't create a tree for a subset (n=2557)

MLIves

Hello all,

I recently ran syntax for a CRT decision tree that worked fine with a file containing 4589 records.

However, when I reran the exact same syntax on a subset of those records (N=2557), no tree was created—just the initial Node with all of the records.

I haven’t found any documentation about this—partly because I’m not sure how to search for it.

I have a similar tree that uses “Any” as the dependent variable and it works fine with the subset.

 

Any ideas why subsetting doesn’t produce a tree for the dependent count variable? 

Possible solutions? 

The dependent nBH_EDPost ranges from a top heavy 0-49 (56% 0).  Would recoding to smaller groups (? 0,1,2,3,4+ (~10% of the total) solve this?

 

In case it helps, here is the syntax.

Most of the variables are dichotomies (0/1), variables starting with ‘n’ are continuous (number of…) variables

TREE nBH_EDPOST BY BH_Pre MED_Pre  IPF_Private_Pre SOLNIT_Pre  

    AnyBHsvcPRE nAnyBHsvcPRE nPrEMPSep CurrentEMPSClient

    EmergencyPsychiatricTimesPrior [s] InpatientPsychiatricTimesLife [s] InpatientPsychiatricTimesPrior [s]

    outofhomepsychiatrictimeslife [s] outofhomepsychiatrictimesprior [s]

    Sex [n]  CalcAge [s] Race4g [n] PrivRes NonEnglish  Region [n]

    IsTANFEligible [n] DCFStatusIntake [n] LivingSituationIntake [n] SchoolAttendancePrior12MonthIntake [n]

    ArrestedPrior12MonthIntake  IsSED 

    CCS_650 CCS_651 CCS_652 CCS_653 CCS_654 CCS_655 CCS_656 CCS_658 CCS_659 

    CCS_660 CCS_661 CCS_663 CCS_670

    CCS_Depr CCS_BIMan CCS_OthMood CCS_OthMood_nonD BH_No654 AutismSD Axis5Intake [s]

    AlcoholordrugProblematDischarge AlcoholDrugsLife AlcoholDrugsPrior 

    OhioScalesFunctioningWorkerIntake [s] AnyTr_IntkDC nTrType [s] AnySchI_IntkDC nSchI [s]

    ReferralSource [n] ResponseModeEMPS [n] MobileContactsDuring [s] OfficevisitsDuring [s]

    NCRef [s] NCRefU  [s] RTM2 [s] CPA [n] PresentingProblemPrimary [n]

    Ohioscalesproblemseverityworkerintake [s]

    CareReferral_2 CareREferral_3 CareReferral_4 CareReferral_5 CareReferral_6

    CareReferral_7 CareReferral_8 CareReferral_9 CareReferral_10 CareReferral_11 CareReferral_12 

    LengthofStay [s] CompTx nSNeed [s] OhioStress [o]

  /TREE DISPLAY=TOPDOWN NODES=STATISTICS BRANCHSTATISTICS=YES NODEDEFS=YES SCALE=AUTO

  /PRINT MODELSUMMARY RISK

  /GAIN SUMMARYTABLE=YES TYPE=[NODE] SORT=DESCENDING CUMULATIVE=NO

  /METHOD TYPE=CRT

  /GROWTHLIMIT MAXDEPTH=AUTO MINPARENTSIZE=100 MINCHILDSIZE=50

  /VALIDATION TYPE=NONE OUTPUT=BOTHSAMPLES.

 

--------------

Melissa Ives

Tues & Wed                                          Mon, Thurs, & Fri

UCONN/DMHAS Research Division     UCONN School of Social Work

410 Capitol Ave., MS #14RSD              38 Prospect Street

Hartford, CT 06106                              Hartford, CT 06103  

860-418-6729 (phone)                         959-200-3684 (phone)

860-418-6692 (fax)                               [hidden email]

[hidden email]

 

 




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
Reply | Threaded
Open this post in threaded view
|

Re: TREE syntax that works doesn't create a tree for a subset (n=2557)

Jon Peck
Most likely it didn't find a significant split or big enough subgroup in the smaller sample.  You could lower the splitting criteria to persuade it to make a tree.

On Tue, Mar 6, 2018 at 10:52 AM, Ives, Melissa L <[hidden email]> wrote:

Hello all,

I recently ran syntax for a CRT decision tree that worked fine with a file containing 4589 records.

However, when I reran the exact same syntax on a subset of those records (N=2557), no tree was created—just the initial Node with all of the records.

I haven’t found any documentation about this—partly because I’m not sure how to search for it.

I have a similar tree that uses “Any” as the dependent variable and it works fine with the subset.

 

Any ideas why subsetting doesn’t produce a tree for the dependent count variable? 

Possible solutions? 

The dependent nBH_EDPost ranges from a top heavy 0-49 (56% 0).  Would recoding to smaller groups (? 0,1,2,3,4+ (~10% of the total) solve this?

 

In case it helps, here is the syntax.

Most of the variables are dichotomies (0/1), variables starting with ‘n’ are continuous (number of…) variables

TREE nBH_EDPOST BY BH_Pre MED_Pre  IPF_Private_Pre SOLNIT_Pre  

    AnyBHsvcPRE nAnyBHsvcPRE nPrEMPSep CurrentEMPSClient

    EmergencyPsychiatricTimesPrior [s] InpatientPsychiatricTimesLife [s] InpatientPsychiatricTimesPrior [s]

    outofhomepsychiatrictimeslife [s] outofhomepsychiatrictimesprior [s]

    Sex [n]  CalcAge [s] Race4g [n] PrivRes NonEnglish  Region [n]

    IsTANFEligible [n] DCFStatusIntake [n] LivingSituationIntake [n] SchoolAttendancePrior12MonthIntake [n]

    ArrestedPrior12MonthIntake  IsSED 

    CCS_650 CCS_651 CCS_652 CCS_653 CCS_654 CCS_655 CCS_656 CCS_658 CCS_659 

    CCS_660 CCS_661 CCS_663 CCS_670

    CCS_Depr CCS_BIMan CCS_OthMood CCS_OthMood_nonD BH_No654 AutismSD Axis5Intake [s]

    AlcoholordrugProblematDischarge AlcoholDrugsLife AlcoholDrugsPrior 

    OhioScalesFunctioningWorkerIntake [s] AnyTr_IntkDC nTrType [s] AnySchI_IntkDC nSchI [s]

    ReferralSource [n] ResponseModeEMPS [n] MobileContactsDuring [s] OfficevisitsDuring [s]

    NCRef [s] NCRefU  [s] RTM2 [s] CPA [n] PresentingProblemPrimary [n]

    Ohioscalesproblemseverityworkerintake [s]

    CareReferral_2 CareREferral_3 CareReferral_4 CareReferral_5 CareReferral_6

    CareReferral_7 CareReferral_8 CareReferral_9 CareReferral_10 CareReferral_11 CareReferral_12 

    LengthofStay [s] CompTx nSNeed [s] OhioStress [o]

  /TREE DISPLAY=TOPDOWN NODES=STATISTICS BRANCHSTATISTICS=YES NODEDEFS=YES SCALE=AUTO

  /PRINT MODELSUMMARY RISK

  /GAIN SUMMARYTABLE=YES TYPE=[NODE] SORT=DESCENDING CUMULATIVE=NO

  /METHOD TYPE=CRT

  /GROWTHLIMIT MAXDEPTH=AUTO MINPARENTSIZE=100 MINCHILDSIZE=50

  /VALIDATION TYPE=NONE OUTPUT=BOTHSAMPLES.

 

--------------

Melissa Ives

Tues & Wed                                          Mon, Thurs, & Fri

UCONN/DMHAS Research Division     UCONN School of Social Work

410 Capitol Ave., MS #14RSD              38 Prospect Street

Hartford, CT 06106                              Hartford, CT 06103  

<a href="tel:(860)%20418-6729" value="+18604186729" target="_blank">860-418-6729 (phone)                         <a href="tel:(959)%20200-3684" value="+19592003684" target="_blank">959-200-3684 (phone)

<a href="tel:(860)%20418-6692" value="+18604186692" target="_blank">860-418-6692 (fax)                               [hidden email]

[hidden email]

 

 




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



--
Jon K Peck
[hidden 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
Reply | Threaded
Open this post in threaded view
|

Re: TREE syntax that works doesn't create a tree for a subset (n=2557)

Rick Oliver
In reply to this post by MLIves
How good is the fit with the whole dataset? Although it seems unlikely, it's possible that with the subset the procedure just doesn't see anything significant going on at the specified node sizes.

On Tue, Mar 6, 2018 at 11:52 AM, Ives, Melissa L <[hidden email]> wrote:

Hello all,

I recently ran syntax for a CRT decision tree that worked fine with a file containing 4589 records.

However, when I reran the exact same syntax on a subset of those records (N=2557), no tree was created—just the initial Node with all of the records.

I haven’t found any documentation about this—partly because I’m not sure how to search for it.

I have a similar tree that uses “Any” as the dependent variable and it works fine with the subset.

 

Any ideas why subsetting doesn’t produce a tree for the dependent count variable? 

Possible solutions? 

The dependent nBH_EDPost ranges from a top heavy 0-49 (56% 0).  Would recoding to smaller groups (? 0,1,2,3,4+ (~10% of the total) solve this?

 

In case it helps, here is the syntax.

Most of the variables are dichotomies (0/1), variables starting with ‘n’ are continuous (number of…) variables

TREE nBH_EDPOST BY BH_Pre MED_Pre  IPF_Private_Pre SOLNIT_Pre  

    AnyBHsvcPRE nAnyBHsvcPRE nPrEMPSep CurrentEMPSClient

    EmergencyPsychiatricTimesPrior [s] InpatientPsychiatricTimesLife [s] InpatientPsychiatricTimesPrior [s]

    outofhomepsychiatrictimeslife [s] outofhomepsychiatrictimesprior [s]

    Sex [n]  CalcAge [s] Race4g [n] PrivRes NonEnglish  Region [n]

    IsTANFEligible [n] DCFStatusIntake [n] LivingSituationIntake [n] SchoolAttendancePrior12MonthIntake [n]

    ArrestedPrior12MonthIntake  IsSED 

    CCS_650 CCS_651 CCS_652 CCS_653 CCS_654 CCS_655 CCS_656 CCS_658 CCS_659 

    CCS_660 CCS_661 CCS_663 CCS_670

    CCS_Depr CCS_BIMan CCS_OthMood CCS_OthMood_nonD BH_No654 AutismSD Axis5Intake [s]

    AlcoholordrugProblematDischarge AlcoholDrugsLife AlcoholDrugsPrior 

    OhioScalesFunctioningWorkerIntake [s] AnyTr_IntkDC nTrType [s] AnySchI_IntkDC nSchI [s]

    ReferralSource [n] ResponseModeEMPS [n] MobileContactsDuring [s] OfficevisitsDuring [s]

    NCRef [s] NCRefU  [s] RTM2 [s] CPA [n] PresentingProblemPrimary [n]

    Ohioscalesproblemseverityworkerintake [s]

    CareReferral_2 CareREferral_3 CareReferral_4 CareReferral_5 CareReferral_6

    CareReferral_7 CareReferral_8 CareReferral_9 CareReferral_10 CareReferral_11 CareReferral_12 

    LengthofStay [s] CompTx nSNeed [s] OhioStress [o]

  /TREE DISPLAY=TOPDOWN NODES=STATISTICS BRANCHSTATISTICS=YES NODEDEFS=YES SCALE=AUTO

  /PRINT MODELSUMMARY RISK

  /GAIN SUMMARYTABLE=YES TYPE=[NODE] SORT=DESCENDING CUMULATIVE=NO

  /METHOD TYPE=CRT

  /GROWTHLIMIT MAXDEPTH=AUTO MINPARENTSIZE=100 MINCHILDSIZE=50

  /VALIDATION TYPE=NONE OUTPUT=BOTHSAMPLES.

 

--------------

Melissa Ives

Tues & Wed                                          Mon, Thurs, & Fri

UCONN/DMHAS Research Division     UCONN School of Social Work

410 Capitol Ave., MS #14RSD              38 Prospect Street

Hartford, CT 06106                              Hartford, CT 06103  

<a href="tel:(860)%20418-6729" value="+18604186729" target="_blank">860-418-6729 (phone)                         <a href="tel:(959)%20200-3684" value="+19592003684" target="_blank">959-200-3684 (phone)

<a href="tel:(860)%20418-6692" value="+18604186692" target="_blank">860-418-6692 (fax)                               [hidden email]

[hidden email]

 

 




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

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