Example of prediction with CART® Regression

A healthcare provider operates a facility that provides substance abuse treatment services. One of the services at the facility is an outpatient detoxification program where a regular course of treatment can last from 1 to 30 days. A team responsible for projecting staffing and supplies wants to study whether they can make better predictions about the length of time a patient uses services. The team uses information collected when the patient enters the program. These variables include demographic information and details about the patient's substance abuse history.

In the example of creating a regression tree, the team chose to examine a regression tree with 7 nodes. The researchers now want to make predictions using this tree.

  1. Complete Example of CART® Regression.
  2. Open the sample dataset LengthOfServicePredictions.MTW.
  3. Ensure that the worksheet that contains the prediction data is active and click the Predict button at the bottom of the regression tree results.
  4. From the drop-down list, select Enter columns of values.
  5. Enter the following values:
    Age at Admission Age at Admission
    Age of First Drug Use Age of First Drug Use
    Arrests in Previous 30 Days Arrests in Previous 30 Days
    Days Waiting for Service Days Waiting for Service
    Previous Treatment Episodes Previous Treatment Episodes
    Years of Education Years of Education
    Other Stimulant Use Other Stimulant Use
    Planned Medication Therapy Planned Medication Therapy
    Psychiatric Condition Psychiatric Condition
    Pregnant Pregnant
    Gender Gender
    Veteran Veteran
    Alcohol Use Alcohol Use
    Cocaine Use Cocaine Use
    Marijuana Use Marijuana Use
    Heroin Use Heroin Use
    Other Opioid Use Other Opioid Use
    PCP Use PCP Use
    Methadone Use Methadone Use
    Other Hallucinogen Use Other Hallucinogen Use
    Methamphetamine Use Methamphetamine Use
    Other Amphetamine Use Other Amphetamine Use
    Benzodiazepine Use Benzodiazepine Use
    Other Tranquilizer Use Other Tranquilizer Use
    Barbituate Use Barbituate Use
    Other Sedative Use Other Sedative Use
    Inhalant Use Inhalant Use
    Non-Prescription Drug Use Non-Prescription Drug Use
    Other Drug Use Other Drug Use
    Intravenous Drug Use Intravenous Drug Use
    Living Arrangements Living Arrangements
    Frequency of Substance Abuse Frequency of Substance Abuse
    Health Insurance Health Insurance
    Marital Status Marital Status
    Ethnicity Ethnicity
    Income Source Income Source
    Primary Ingestion Route of Sub Primary Ingestion Route of Sub
    Self-Help Attendance Self-Help Attendance
    Source of Payment Source of Payment
    Race Race
    Employment Status Employment Status
    Referral Source Referral Source
    Primary Substance of Abuse Primary Substance of Abuse
    DSM Diagnosis DSM Diagnosis
  6. Click OK.

Interpret the results

Minitab uses the tree that corresponds to the Predict button to estimate the response values from the prediction data. Because the prediction values are stored in columns, the results are in the worksheet.

The predictions allocate the 7 observations into 5 terminal nodes: 1, 5, 8, 12, and 17. The following table shows the means for these terminal nodes, rounded to the nearest hundredth of a day:
Terminal node Mean
1 5.85
5 10.17
8 16.69
12 20.11
17 27.56
By using this site you agree to the use of cookies for analytics and personalized content.  Read our policy