2. experience sampling
• a form of moment-to-moment
data collection
• increased ecological validity
• minimise retrospective bias
• participant burden
• different kinds of questions
3. experience sampling
• a form of moment-to-moment
data collection
• increased ecological validity
• minimise retrospective bias
• participant burden
• different kinds of questions
4. experience sampling
• a form of moment-to-moment
data collection
• increased ecological validity
• minimise retrospective bias
• participant burden
• different kinds of questions
5. experience sampling
• a form of moment-to-moment
data collection
• increased ecological validity
• minimise retrospective bias
• participant burden
• different kinds of questions
6. experience sampling
• a form of moment-to-moment
data collection
• increased ecological validity
• minimise retrospective bias
• participant burden
• different kinds of questions
7. Jeffrey S. Simons, Raluca M. Gaher, Matthew N.I. Oliver, Jacqueline A. Bush, Marc A. Palmer
An Experience Sampling Study of Associations between Affect
and Alcohol Use and Problems among College Students
example
8. a quick note; I am focused on self report
studies but passive data collection is also
possible
9. design considerations
• appropriate measurement
resolution
couple satisfaction
• event-based versus interval-
based response cues
• sample size and power
• engaging participants
• response medium
blood glucose monitoring
10. design considerations
• appropriate measurement
resolution
event-based
• event-based versus interval-
based response cues
• sample size and power
• engaging participants
• response medium interval-based
11. design considerations
• appropriate measurement
resolution
!
Mass & Hox (2005) Sufficient
Sample Sizes for Multilevel
Modeling
• rough rule of thumb: 50
individuals
• although power depends on
many factors and is often
most usefully estimated
based on power analysis
• event-based versus interval-
based response cues
• sample size and power
• engaging participants
• response medium
12. design considerations
• appropriate measurement
resolution
• event-based versus interval-
based response cues
• sample size and power
• engaging participants
• response medium
• honorarium
• usability
• length / frequency
• feedback
13. design considerations
• appropriate measurement
resolution
• event-based versus interval-
based response cues
• sample size and power
• engaging participants
• response medium
14. design considerations
• appropriate measurement
resolution
• event-based versus interval-
based response cues
• sample size and power
• engaging participants
• response medium
15. design considerations
• appropriate measurement
resolution
• event-based versus interval-
based response cues
• sample size and power
• engaging participants
• response medium
16. design considerations
• appropriate measurement
resolution
• event-based versus interval-
based response cues
• sample size and power
• engaging participants
• response medium
17. design considerations
• appropriate measurement
resolution
• event-based versus interval-
based response cues
• sample size and power
• engaging participants
• response medium
PDAs
18. design considerations
• appropriate measurement
resolution
• event-based versus interval-
based response cues
• sample size and power
• engaging participants
• response medium
web surveys
19. resources for optimising
web forms for mobile
• detecting whether participant is using mobile
• optimise webpage for iOS
20. design considerations
• appropriate measurement
resolution
• event-based versus interval-
based response cues
• sample size and power
• engaging participants
• response medium
mobile application
21. design considerations
• appropriate measurement
resolution
• event-based versus interval-
based response cues
• sample size and power
• engaging participants
• response medium
mobile application
22. analysis
• main difference between ‘regular’ analysis and
analysis of ESM data is the hierarchical structure of
the data
level 1: time points
23. analysis
• main difference between ‘regular’ analysis and
analysis of ESM data is the hierarchical structure of
the data
{
{
{
{
{
{
level 1: time points
level 2: individuals
24. analysis
• multilevel modeling (MLM) addresses the lack of
independence between the observations
• can also use regression with robust standard errors
• in addition, MLM opens up some possibilities for
some novel questions not so easily answered in
single level analyses
25. example
• using ESM to study risky single occasion drinking
• presentation that follows is mostly visual, do not
take the diagrams too literally. for more
comprehensive / technical overview of MLM as
applied to ESM data please see
• Intensive Longitudinal Methods: An Introduction to Diary and
Experience Sampling Research
• Models for intensive longitudinal data
26. intercept only model
risky drinking
fun seeking
level 1 variable
level 2 variable
clustering variable = participant id
positive moodeveningpositive mood
27. intercept only
• Intraclass correlation (degree of variance explained
in the outcome variable by the clustering / nesting
variable)
28. intercept only
• Intraclass correlation (degree of variance explained
in the outcome variable by the clustering / nesting
variable)
29. rsod on positive mood
risky drinking
fun seeking
level 1 variable
level 2 variable
clustering variable = participant id
positive mood
evening
positive mood
30. level 1 variables
• level 1 variables actually capture two sources of
variance:
• within participant variation (e.g., fluctuations around
an individual’s average level of mood)
• between participant variation (e.g., individual
differences in level of positive mood)
• these are often usefully represented using separate
variables in the model
• achieved by person mean centring
31. level 1 variables
level 1 positive mood = score - person’s mean
!
!
level 2 positive mood = individual’s average across time
points
32. level 1 variables
• fixed component of an effect
• average relationship between variables for all
participants
• e.g., on average, how does positive mood relate to
drinking?
!
• random component
• between participant variance in relationship
• e.g., how much variation is there in the relationship
between positive mood and drinking? Does positive
mood more strongly associate with drinking for some
participants compared to others?
33. rsod on positive mood
risky drinking
fun seeking
level 1 variable
level 2 variable
clustering variable = participant id
positive mood
evening
positive mood
34. level 2 moderators
• can we explain variation in level 1 relationships
using level 2 variables?
• E.g., does an individual’s fun seeking explain
variation in the relationship between positive
mood and drinking?
39. photo credits
Couple photo: https://flic.kr/p/4SDwWz !
"Couple in Covent Garden" by Mark Hillary (https://www.flickr.com/photos/markhillary/)!
!
Diabetes photo!
"My "kit"" by Jessica Merz (https://www.flickr.com/photos/jessicafm/)!
!
Alcohol photo!
"Alcohol and Ulcerative Colitis" by Kimery Davis (https://www.flickr.com/photos/117025355@N05/)!
!
Timer photo!
"Microwave Timer" by Pascal (https://www.flickr.com/photos/pasukaru76/)!
!
PDA photo!
"I Used To Be Cool..." by H. Michael Karshis (https://www.flickr.com/photos/hmk/)!
!
Piecewise regression graph
http://www3.nd.edu/~rwilliam/stats2/l61.pdf