4. Background
• Used 1983 and 2001 data
from the NSF S&T surveys
• Found that images of
scientists driven by:
• Year (2001 = better)
• Demographics
• Faith vs. science
6. Current Project • Not feasible to include 1983 data in new analysis
• As of 2006, NSF S&T survey is
part of General Social Survey (GSS)
• Not much overlap between 1983 and 2012
surveys
• Losh (2010) include a single “images” of scientist index
• Initial analysis suggested that available
items do not scale into a single index
• Also potential for additional predictor variables aimed
at understanding origins of views about scientists
Two questions:
• Is there evidence of a change by descriptive statistics?
(While being wary of survey mode change)
• Did predictors of views change by year?
The data and the analysis …
• 2001: n = ~1,000-1,200 (RDD telephone)
• 2012: n = ~400-900 (Face-to-Face)
Step 1. Mean comparisons using t-tests/ANOVA
Step 2. GLM Model with interactions by survey year
8. Challenges
Descriptive Statistics and 2001-2012 Comparisons for Criterion Variables
Slight increase
in positive
perceptions
Slight increase
in negative
perceptions
9. Challenges
GLM tests of between subject effects for views about scientists
Year had very
little impact on
the underlying
relationships
Note: These are F-scores, NOT parameter estimates!
10. Challenges
GLM tests of between subject effects for views about scientists
Women were
slightly less likely to
have negative views
(b = -.12 and -.11)
Note: These are F-scores, NOT parameter estimates!
11. Challenges
GLM tests of between subject effects for views about scientists
Young people and
those who took
more math and
science courses
tended to have
fewer negative
views
Note: These are F-scores, NOT parameter estimates!
12. Challenges
GLM tests of between subject effects for views about scientists
Those with more
interest are more
likely to hold
positive images
(b = .15, but
less so in 2001)
Note: These are F-scores, NOT parameter estimates!
13. Challenges
GLM tests of between subject effects for views about scientists
Small negative
relationship
between
knowledge and
perceptions
(b = -.02 to -.04)
Note: These are F-scores, NOT parameter estimates!
14. Challenges
GLM tests of between subject effects for views about scientists
No museum visit
or lots of museum
visits associated
with seeing more
danger in 2012
(i.e., non-linear)
Note: These are F-scores, NOT parameter estimates!
15. Challenges
GLM tests of between subject effects for views about scientists
Indicating that
newspaper were
the primary
source of S&T was
associated with
lower perceptions
of danger and
working alone
Note: These are F-scores, NOT parameter estimates!
16. Challenges
GLM tests of between subject effects for views about scientists
Indicating that
newspaper were
the primary
source of S&T was
associated with
lower perceptions
of danger and
working alone
(b = -.18 and -.20)
Note: These are F-scores, NOT parameter estimates!
17. Challenges
GLM tests of between subject effects for views about scientists
Very limited
variance
explained
Note: These are F-scores, NOT parameter estimates!
18. Discussion
• Factors associated with scientist views include:
• Age and gender
• Experience/Knowledge of science
BUT …
• NSF S&T Survey is meant to be a key source
of S&T knowledge and attitude data.
AND …
• Models based on available questions explain
limited variance in scientist perceptions
SO …
• We need to encourage NSF to continue to think
about the questions included in the S&T survey
• Better communication variables, including
exposure/attention to various sources of science
content, as well as interpersonal discussion
• Consistent issue specific and general attitude measures
Next steps …
• Redo the analysis in MPlus using multi-group
modelling approach
• Build in criterion variables related to impact of views