Contributions of negative experiences

1
Contributions of negative experiences

• The contributions of sexual harassment,
  aggressive experiences, and work discrimination
  to PTSD and job withdrawal intentions

2
Contributions of negative experiences
Wayne C. Lee The Boeing Company and the
University of Illinois at Urbana-Champaign Alayne
J. Ormerod, Maggie E. Reed, and Jorja
Jamison University of Illinois at Urbana-Champaign
3
Overview

• Extensive work conducted on negative experiences
  in the workplace
• Typically each class of negative experiences
  individually
• Can we begin to understand the interplay between
  them?

4
Current study

• Goal Explore the relative contributions of three
  classes of negative experiences
• Data pooled from two similar surveys
• Use various methods to examine the contributions
  of these negative experiences

5
Sample

• Two sources
• Self-administered, paper-and-pencil survey given
  to 737 active duty military personnel
• On-line survey, 211 active duty military
  personnel
• Similar survey content
• Current study was not intention of surveys

6
Measures

• Sexual harassment
• Sexual experiences Questionnaire (SEQ Fitzgerald
  et al., 1995, Stark et al., in press)
• Aggressive experiences in the workplace
• Aggressive Experiences Scale-Target (AES-Target
  Glomb, 2002)
• Workplace discrimination (Scarville et al., 1999)

7
Outcomes in question

• Post-traumatic stress disorder (symptoms)
• PTSD Checklist (PCL Weathers, Litz, Herman,
  Huska, Keane (1993)
• Turnover intentions
• Single item

8
Method

• Traditional regression plus dominance analyses
• Classification and regression trees (Breiman,
  Friedman, Olshen, Stone, 1984)
• Computer based, data-mining technique using
  decision trees

9
CART in a nutshell

• Can be used to predict any number of outcomes
  based on a set of predictor variables
• Examines all possible splits of the data based on
  individual cut-off scores of input variables
• Binary splits form the nodes of tree

10
CART continued

• Nodes are arranged beginning with the most
  discriminating
• CART then prunes the tree using a number of
  heuristics
• CART can determine rates of misclassification

11
Why CART?

• A number of benefits
• Separate paths can lead to the same outcome
• Predictor variables may be reused
• CART is nonparametric
• May exclude predictor variables
• Has been shown to tolerate outliers

12
Examining PTSD

• Predictors SEQ, Workplace Aggression, Work
  Discrimination
• Outcome PTSD
• Determined by a counting algorithm using the PCL
  scale scores
• 3 on a 5-point scale for sufficient symptoms
  (items)
• N211

13
CART results

• 7 trees grown
• Best tree 57.14 hit rate, 24.66 false positive
  rate (validation sub-sample)
• Workplace aggression and SEQ identified as the
  most valuable predictors

14
CART modeling of PTSD likelihood
15
CART tree
16
Logistic regression vs CART
17
Examining turnover intentions

• Predictors SEQ, Workplace Aggression, Work
  Discrimination
• Outcome Turnover intentions
• Methods CART, logistic regression
• N630

18
CART results

• Results not as good as with PTSD
• 23 trees
• Best tree 52.33 hit rate, 36.64 false positive
  rate (validation sample)
• Work discrimination identified as the most
  valuable predictor

19
CART tree
20
Logistic regression vs CART
21
Additional analyses

• Dominance analysis
• Mirrors findings with CART
• Most influential predictors

22
Summary of findings

• Moderate evidence of an interplay between these
  three forms of negative experiences
• CART can be a beneficial tool in mapping out
  these relationships

23
Room for improvement?

• Low base rate for phenomena
• Measures could be improved
• Low N

24
More info

• wlee_at_s.psych.uiuc.edu
• aormerod_at_uiuc.edu
• Thanks!