1 / 25

# Microsimulation Collection Project - PowerPoint PPT Presentation

Microsimulation Collection Project. Kristen Couture Yves Bélanger Elisabeth Neusy Marcelle Tremblay. Outline. Overview Models created prior to Simulation Call Outcomes Call Duration Simulation Model SAS Simulation Studio program overview Aspects of Simulation Some Early Results

I am the owner, or an agent authorized to act on behalf of the owner, of the copyrighted work described.

Microsimulation Collection Project

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.

- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -

## Microsimulation Collection Project

Kristen Couture

Yves Bélanger

Elisabeth Neusy

Marcelle Tremblay

### Outline

• Overview

• Models created prior to Simulation

• Call Outcomes

• Call Duration

• Simulation Model

• SAS Simulation Studio program overview

• Aspects of Simulation

• Some Early Results

• Conclusions and Future Work

### Overview

• What are we trying to do?

• Construct a simulation model that will represent the CATI collection process using SAS Simulation Studio

• Why are we doing this?

• To attempt to find ways to optimise collection activities that will make collection more efficient within a controlled environment

### Overview

• Questions we are trying to answer:

• What effect do time slices have on the collection process?

• How does the distribution of interviewers affect collection?

• How does the introduction of a cap on calls affect the overall response rate?

### Steps to Building Simulation

Simulation

Collection Parameters

### Modelling Call Outcomes

• 5 outcomes: Unresolved, Out of Scope, Refusal, Other Contact, Respondent

• Modelled Using Multinomial Logistic Regression and CSGVP 2004 BTH

• 7 parameters entered into the model

i = 1..n

j = 1..k

Parameters Data Set

### Modelling Call Outcomes

• Calculate probability for each possible call outcome using estimated betas and collection parameters

### Modelling Call Duration

• Use 2004 CSGVP BTH

• Draw histograms for each outcome

• Use Probability Plots to Determine Distribution and Parameters

Response Histogram

Normal Probability Plot

D

U

R

A

T

I

O

N

P

E

R

C

E

N

T

Normal Percentiles

Call Duration

### Aspects of Simulation

• Consists of…

• Input: user enters parameters for model

• Clock: Creates parameters from simulation clock

• Queue: calls wait to be interviewed

• Call Center: calls are made, outcome and duration of call is simulated

• Interviewer Agenda: change # of interviewers

• Time Slices (in progress): maximum number of attempts implemented for each time slice

• Output: BTH file

### Input

• Allows user to enter parameters via SAS Data Sets

Parameters Data Set

Time Slice Data Sets

### Clock

• Creates Time Parameters including Evening, Weekend, PM, and Time Slices by reading the current simulation time

### Queuing System

• Cases are created and enter a queue waiting to be interviewed

### Determining Call Outcome

• Determines Call Outcome:

• Unresolved

• Out of Scope

• Other Contact

• Refusal

• Respondent

### Call Center

• Call is sent to Call Center where it is interviewed

### Call Center

• User can change the number of interviewers during a specified time period

### Finalizing Cases

• Outcome of Out of Scope or Respondent

• Reached Cap on Calls

• Residential: 20

• Unknown: 5

• Number of Refusals=3

• Output is created in terms of SAS data set

### Simulation Example

• Create 10,000 cases and run the simulation for 30 days of collection

• Interviewers:

• Shift 1 (9am-12pm) : 10

• Shift 2 (12pm-5pm) : 10

• Shift 3 (5pm-9pm) : 10

*Note: No time slices in this example

### Diagnostics

Finalized Cases and Response Rate

Distribution of Outcome Codes

### Diagnostics

Last Call Outcome

Last Call Outcome by Original Residential Status

### Changing Parameters

Effect on changing the number of interviewers and days of collection

### Conclusions

• Allows user to enter parameters into model

• Reproduce results similar to CSGVP 2004

• Create a BTH file

• Change parameters and look at the effect

### Future Work

• Improve the model by adding more parameters

• Produce results with time slices implemented to model to measure impact

• Add attributes to the interviewers such as English/French/bilingual and Senior/Junior

• Rearrange the cases in the queue so that they will be pre-empted at best time to call