Secondary results
Straight-lining
Induction task plots suggest a lot of anchoring around the 50%
marker, and debriefing suggests many participants thought it was a
strange task since the test features were odd/bizarre. Are these just
from participants who were straightlining through all the test
items?
3 out of 507 included participants (0.59%) answered 50% to all
test questions.
5 out of 507 included participants (0.99%) answered 48-52% to all
test questions, a looser criterion.
Since there were only a few participants who consistently
straightlined, these participants were not excluded from analyses.
Pairwise condition comparisons
We can make pairwise comparisons between conditions
using two-sample Kolmogorov–Smirnov tests, with Bonferroni correction
for number of tests run (105). The Kolmogorov–Smirnov test compares the
cumulative distributions of two samples and returns a statistic
D that reflects the maximum difference between the two
distributions, as well as a p-value for the test.
One quick and dirty way to think about these results is to look at
how often pairs of conditions with the same (boxed) versus
different (not in box) number of generics are significantly
different from each other, and to do the same for pairs of conditions
with same (boxed) versus different (not in box)
proportions of generics.
Note, there are way way more pairwise comparisons comparing
different numbers or proportions of generics than same
numbers or proportion of generics, so this is a bit of a lopsided
comparison.
If say number of generics matters, we would expect to see that the
distribution of prevalence ratings rarely differ when comparing pairs
that are the same number of generics, and differ much more when
comparing pairs that are different number of generics.
If say proportion of generics matters, we would expect to see that
the distribution of prevalence ratings rarely differ when comparing
pairs that are the same proportion of generics, and differ much
more when comparing pairs that are different proportion of
generics.
| same_num_generics |
sig_corr_tests |
total_tests |
prop_sig_corr |
| FALSE |
46 |
85 |
54.1% |
| TRUE |
4 |
20 |
20.0% |
| same_prop_generics |
sig_corr_tests |
total_tests |
prop_sig_corr |
| FALSE |
44 |
78 |
56.4% |
| TRUE |
0 |
13 |
0.0% |
| NA |
6 |
14 |
42.9% |
Study 6 conditions (replication)
Three of the conditions (baseline, 0/16, 16/16) in this study
(n = 32-35/condition) are replications of the conditions
(baseline, specific, generic conditions) in Study 6 (n =
90-99/condition). In Study 6, prevalence was rated higher in the generic
condition than the baseline condition than the specific condition,
respectively.
## # A tibble: 3 × 2
## condition count
## <fct> <int>
## 1 0/16 35
## 2 baseline 35
## 3 16/16 35
After subsetting to the Study 6 conditions, we do replicate the main
effect of condition (\(\chi\)(2) =
12.61, p = .0018). The 16/16 condition reported marginally
higher prevalence judgments compared to baseline (z = 2.19,
p = .085), and significantly higher compared to 0/16 condition
(z = 3.51, p = .0013).
Unlike in Study 6, the baseline and 0/16 conditions were not
statistically different from each other (z = 1.33, p =
.56).
# same analysis as study 6
model <-
glmmTMB(prevalence ~ condition + (1|participant) + (1|test_feature),
data = data_study_6,
family = beta_family(link = "logit"))
model %>%
Anova()
## Analysis of Deviance Table (Type II Wald chisquare tests)
##
## Response: prevalence
## Chisq Df Pr(>Chisq)
## condition 12.612 2 0.001825 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
model %>%
emmeans("condition") %>%
pairs(adjust = "bonferroni") %>%
summary()
## contrast estimate SE df z.ratio p.value
## (16/16) - baseline 0.435 0.199 Inf 2.191 0.0853
## (16/16) - (0/16) 0.698 0.199 Inf 3.516 0.0013
## baseline - (0/16) 0.263 0.199 Inf 1.326 0.5550
##
## Results are given on the log odds ratio (not the response) scale.
## P value adjustment: bonferroni method for 3 tests
Extreme conditions only
Subsetting to the extreme conditions where the proportion of generics
was 0% or 100% (0/4, 0/8, 0/12, 0/16, 4/4, 8/8, 12/12, 16/16).
## # A tibble: 8 × 2
## condition count
## <chr> <int>
## 1 8/8 35
## 2 4/4 33
## 3 16/16 35
## 4 12/12 35
## 5 0/8 32
## 6 0/4 32
## 7 0/16 35
## 8 0/12 33
When each entered on their own as lone predictors, the raw number of
generics (z = 5.33, p < .001), raw number of
specifics (z = -5.13, p < .001), and the proportion
of generics to specifics (z = 5.71, p < .001) were
each significant in predicting prevalence judgments.
When pitted against each other in the same model, the
proportion of generics to specifics remains a
significant predictor of prevalence judgments (z = 2.21,
p = .027), while the raw number of generics
does not (z = 1.00, p = .32).
# raw number of generics
glmmTMB(prevalence ~ num_generics + (1|participant) + (1|test_feature),
data = data_extreme_only,
family = beta_family(link = "logit")) %>%
summary()
## Family: beta ( logit )
## Formula:
## prevalence ~ num_generics + (1 | participant) + (1 | test_feature)
## Data: data_extreme_only
##
## AIC BIC logLik deviance df.resid
## -2966.6 -2934.8 1488.3 -2976.6 4315
##
## Random effects:
##
## Conditional model:
## Groups Name Variance Std.Dev.
## participant (Intercept) 0.6360 0.7975
## test_feature (Intercept) 0.1861 0.4314
## Number of obs: 4320, groups: participant, 270; test_feature, 16
##
## Dispersion parameter for beta family (): 2.68
##
## Conditional model:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -0.665090 0.127314 -5.224 0.000000175 ***
## num_generics 0.045591 0.008563 5.325 0.000000101 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
# raw number of specifics
glmmTMB(prevalence ~ num_specifics + (1|participant) + (1|test_feature),
data = data_extreme_only,
family = beta_family(link = "logit")) %>%
summary()
## Family: beta ( logit )
## Formula:
## prevalence ~ num_specifics + (1 | participant) + (1 | test_feature)
## Data: data_extreme_only
##
## AIC BIC logLik deviance df.resid
## -2964.8 -2932.9 1487.4 -2974.8 4315
##
## Random effects:
##
## Conditional model:
## Groups Name Variance Std.Dev.
## participant (Intercept) 0.6407 0.8005
## test_feature (Intercept) 0.1862 0.4315
## Number of obs: 4320, groups: participant, 270; test_feature, 16
##
## Dispersion parameter for beta family (): 2.68
##
## Conditional model:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -0.211682 0.126739 -1.670 0.0949 .
## num_specifics -0.044006 0.008584 -5.127 0.000000295 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
# proportion of generics
glmmTMB(prevalence ~ prop_generics + (1|participant) + (1|test_feature),
data = data_extreme_only,
family = beta_family(link = "logit")) %>%
summary()
## Family: beta ( logit )
## Formula:
## prevalence ~ prop_generics + (1 | participant) + (1 | test_feature)
## Data: data_extreme_only
##
## AIC BIC logLik deviance df.resid
## -2970.5 -2938.6 1490.2 -2980.5 4315
##
## Random effects:
##
## Conditional model:
## Groups Name Variance Std.Dev.
## participant (Intercept) 0.6260 0.7912
## test_feature (Intercept) 0.1861 0.4314
## Number of obs: 4320, groups: participant, 270; test_feature, 16
##
## Dispersion parameter for beta family (): 2.68
##
## Conditional model:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -0.7261 0.1301 -5.583 0.0000000237 ***
## prop_generics 0.5792 0.1014 5.712 0.0000000112 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
# raw number vs proportion of generics
glmmTMB(prevalence ~ num_generics + prop_generics + (1|participant) + (1|test_feature),
data = data_extreme_only,
family = beta_family(link = "logit")) %>%
summary()
## Family: beta ( logit )
## Formula:
## prevalence ~ num_generics + prop_generics + (1 | participant) +
## (1 | test_feature)
## Data: data_extreme_only
##
## AIC BIC logLik deviance df.resid
## -2969.5 -2931.3 1490.7 -2981.5 4314
##
## Random effects:
##
## Conditional model:
## Groups Name Variance Std.Dev.
## participant (Intercept) 0.6235 0.7896
## test_feature (Intercept) 0.1861 0.4314
## Number of obs: 4320, groups: participant, 270; test_feature, 16
##
## Dispersion parameter for beta family (): 2.68
##
## Conditional model:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -0.72614 0.12999 -5.586 0.0000000232 ***
## num_generics 0.01585 0.01589 0.997 0.319
## prop_generics 0.41931 0.18959 2.212 0.027 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Mixed conditions only
Mixed conditions (4/8, 4/12, 4/16, 8/12, 8/16, 12/16) have heightened
contrast between generic and specific statements, since participants see
both types of statements.
## # A tibble: 6 × 2
## condition count
## <chr> <int>
## 1 8/16 33
## 2 8/12 34
## 3 4/8 35
## 4 4/16 34
## 5 4/12 34
## 6 12/16 32
After subsetting the sample to just mixed conditions (n = 202 total),
neither raw numbers of generics, raw numbers of specifics, nor the
proportion of generics to specifics were statistically significant in
predicting prevalence judgments on their own (ps > .31).
# raw number of generics
glmmTMB(prevalence ~ num_generics + (1|participant) + (1|test_feature),
data = data_mixed_only,
family = beta_family(link = "logit")) %>%
summary()
## Family: beta ( logit )
## Formula:
## prevalence ~ num_generics + (1 | participant) + (1 | test_feature)
## Data: data_mixed_only
##
## AIC BIC logLik deviance df.resid
## -2163.1 -2132.7 1086.5 -2173.1 3227
##
## Random effects:
##
## Conditional model:
## Groups Name Variance Std.Dev.
## participant (Intercept) 0.7237 0.8507
## test_feature (Intercept) 0.2077 0.4557
## Number of obs: 3232, groups: participant, 202; test_feature, 16
##
## Dispersion parameter for beta family (): 2.92
##
## Conditional model:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -0.55505 0.19065 -2.911 0.0036 **
## num_generics 0.02149 0.02115 1.016 0.3096
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
# raw number of specifics
glmmTMB(prevalence ~ num_specifics + (1|participant) + (1|test_feature),
data = data_mixed_only,
family = beta_family(link = "logit")) %>%
summary()
## Family: beta ( logit )
## Formula:
## prevalence ~ num_specifics + (1 | participant) + (1 | test_feature)
## Data: data_mixed_only
##
## AIC BIC logLik deviance df.resid
## -2162.6 -2132.2 1086.3 -2172.6 3227
##
## Random effects:
##
## Conditional model:
## Groups Name Variance Std.Dev.
## participant (Intercept) 0.7257 0.8519
## test_feature (Intercept) 0.2077 0.4557
## Number of obs: 3232, groups: participant, 202; test_feature, 16
##
## Dispersion parameter for beta family (): 2.92
##
## Conditional model:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -0.31360 0.19074 -1.644 0.100
## num_specifics -0.01495 0.02093 -0.714 0.475
# proportion of generics
glmmTMB(prevalence ~ prop_generics + (1|participant) + (1|test_feature),
data = data_mixed_only,
family = beta_family(link = "logit")) %>%
summary()
## Family: beta ( logit )
## Formula:
## prevalence ~ prop_generics + (1 | participant) + (1 | test_feature)
## Data: data_mixed_only
##
## AIC BIC logLik deviance df.resid
## -2163.0 -2132.6 1086.5 -2173.0 3227
##
## Random effects:
##
## Conditional model:
## Groups Name Variance Std.Dev.
## participant (Intercept) 0.7238 0.8508
## test_feature (Intercept) 0.2077 0.4557
## Number of obs: 3232, groups: participant, 202; test_feature, 16
##
## Dispersion parameter for beta family (): 2.92
##
## Conditional model:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -0.5938 0.2221 -2.673 0.00752 **
## prop_generics 0.3626 0.3620 1.002 0.31651
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
# raw number vs proportion of generics
glmmTMB(prevalence ~ num_generics * prop_generics + (1|participant) + (1|test_feature),
data = data_mixed_only,
family = beta_family(link = "logit")) %>%
summary()
## Family: beta ( logit )
## Formula:
## prevalence ~ num_generics * prop_generics + (1 | participant) +
## (1 | test_feature)
## Data: data_mixed_only
##
## AIC BIC logLik deviance df.resid
## -2159.1 -2116.6 1086.6 -2173.1 3225
##
## Random effects:
##
## Conditional model:
## Groups Name Variance Std.Dev.
## participant (Intercept) 0.7234 0.8505
## test_feature (Intercept) 0.2077 0.4557
## Number of obs: 3232, groups: participant, 202; test_feature, 16
##
## Dispersion parameter for beta family (): 2.92
##
## Conditional model:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -0.549962 0.603784 -0.911 0.362
## num_generics 0.004829 0.129493 0.037 0.970
## prop_generics 0.125824 1.097975 0.115 0.909
## num_generics:prop_generics 0.011343 0.180743 0.063 0.950
# raw number of generics vs raw number of specifics
glmmTMB(prevalence ~ num_generics + num_specifics + (1|participant) + (1|test_feature),
data = data_mixed_only,
family = beta_family(link = "logit")) %>%
summary()
## Family: beta ( logit )
## Formula:
## prevalence ~ num_generics + num_specifics + (1 | participant) +
## (1 | test_feature)
## Data: data_mixed_only
##
## AIC BIC logLik deviance df.resid
## -2161.1 -2124.7 1086.6 -2173.1 3226
##
## Random effects:
##
## Conditional model:
## Groups Name Variance Std.Dev.
## participant (Intercept) 0.7234 0.8506
## test_feature (Intercept) 0.2077 0.4557
## Number of obs: 3232, groups: participant, 202; test_feature, 16
##
## Dispersion parameter for beta family (): 2.92
##
## Conditional model:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -0.495465 0.304990 -1.624 0.104
## num_generics 0.018517 0.024244 0.764 0.445
## num_specifics -0.005996 0.023961 -0.250 0.802
Training features
Participants saw some subset of 16 training features - anywhere from
0 to 16 of the full list of 16 - with subsets randomly selected.
Theoretically, the particular training features received might have
an effect on prevalence judgments. Of note, the conditions with fewer
total training utterances contain significant variability in which
training features were seen, while the conditions with more total
training utterances have less variability (with all participants in any
of the 16 utterance conditions all seeing the same features).
Random effects to account for: * different test features may have
different prevalences –> random intercepts per test feature *
different participants may rate prevalence differently –> random
intercepts per participant * total number of utterances heard may
correlate with variability between participants in prevalence ratings,
since conditions with fewer total utterances provided different training
features more variability between-participants in the training –>
random slopes of __ per total number of utterances * different training
features may increase or decrease the overall prevalence
Are the effects above robust after accounting for additional variance
in the training structure?
The proportion of generics to specifics was marginal in predicting
prevalence judgments (z = 1.88, p = .061), while the
number of generics was not significant (z = 0.94, p =
.35) in a beta regression with random slopes for the total utterances
heard, in addition to random intercepts per participant and test
feature.
## Family: beta ( logit )
## Formula:
## prevalence ~ num_generics + prop_generics + (num_generics | total_utt) +
## (prop_generics | total_utt) + (1 | participant) + (1 | test_feature)
## Data: data_tidy_dummy
##
## AIC BIC logLik deviance df.resid
## NA NA NA NA 7588
##
## Random effects:
##
## Conditional model:
## Groups Name Variance
## total_utt (Intercept) 0.005708217899479856721756121373800851870328188
## num_generics 0.000003051523744080734600953508026122129592750
## total_utt.1 (Intercept) 0.000000000000000000000000000000000000000007931
## prop_generics 0.000000000000000000000000000000000006806329817
## participant (Intercept) 0.654452908043517522784782158851157873868942261
## test_feature (Intercept) 0.198675480804334836371083383710356429219245911
## Std.Dev. Corr
## 0.075552749119273329703184
## 0.001746861111846255074087 1.00
## 0.000000000000000000002816
## 0.000000000000000002608894 1.00
## 0.808982637665059844955806
## 0.445730278087920372964703
## Number of obs: 7600, groups: total_utt, 4; participant, 472; test_feature, 16
##
## Dispersion parameter for beta family (): 2.76
##
## Conditional model:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -0.71438 0.13415 -5.325 0.000000101 ***
## num_generics 0.01678 0.01786 0.939 0.3476
## prop_generics 0.39146 0.20872 1.876 0.0607 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
With 16 random intercepts for whether each of the training features
was seen or not, the proportion of generics to specifics is marginal in
predicting prevalence judgments (z = 1.81, p = .0070)
and the raw number of generics is not significant (z = 1.19,
p = .23).
## Family: beta ( logit )
## Formula:
## prevalence ~ num_generics + prop_generics + (1 + num_generics +
## prop_generics || participant) + (1 + num_generics + prop_generics ||
## test_feature) + (1 | trained_babies_blankets) + (1 | trained_bounce_ball_head) +
## (1 | trained_can_flip_air) + (1 | trained_chase_shadows) +
## (1 | trained_climb_fences) + (1 | trained_dont_like_icecream) +
## (1 | trained_dont_like_mud) + (1 | trained_draw_stars_knees) +
## (1 | trained_eat_flowers) + (1 | trained_flap_arms_happy) +
## (1 | trained_freckles_feet) + (1 | trained_hop_puddles) +
## (1 | trained_like_sing) + (1 | trained_scared_ladybugs) +
## (1 | trained_sleep_trees) + (1 | trained_stripes_hair)
## Data: data_tidy_dummy
##
## AIC BIC logLik deviance df.resid
## NA NA NA NA 7574
##
## Random effects:
##
## Conditional model:
## Groups Name Variance Std.Dev. Corr
## participant (Intercept) 0.53384904752 0.7306497
## num_generics 0.00182334595 0.0427007 0.00
## prop_generics 0.06582384614 0.2565616 0.00 0.00
## test_feature (Intercept) 0.17239813713 0.4152085
## num_generics 0.00020475765 0.0143094 0.00
## prop_generics 0.04975536048 0.2230591 0.00 0.00
## trained_babies_blankets (Intercept) 0.00000001462 0.0001209
## trained_bounce_ball_head (Intercept) 0.00000012714 0.0003566
## trained_can_flip_air (Intercept) 0.00000001751 0.0001323
## trained_chase_shadows (Intercept) 0.00000040417 0.0006357
## trained_climb_fences (Intercept) 0.00000001393 0.0001180
## trained_dont_like_icecream (Intercept) 0.00000004709 0.0002170
## trained_dont_like_mud (Intercept) 0.00000005112 0.0002261
## trained_draw_stars_knees (Intercept) 0.00000003362 0.0001834
## trained_eat_flowers (Intercept) 0.00000034332 0.0005859
## trained_flap_arms_happy (Intercept) 0.00000001857 0.0001363
## trained_freckles_feet (Intercept) 0.00000001538 0.0001240
## trained_hop_puddles (Intercept) 0.00000001990 0.0001411
## trained_like_sing (Intercept) 0.00353389795 0.0594466
## trained_scared_ladybugs (Intercept) 0.00000001510 0.0001229
## trained_sleep_trees (Intercept) 0.00000001317 0.0001148
## trained_stripes_hair (Intercept) 0.00000457870 0.0021398
## Number of obs: 7600, groups:
## participant, 472; test_feature, 16; trained_babies_blankets, 2; trained_bounce_ball_head, 2; trained_can_flip_air, 2; trained_chase_shadows, 2; trained_climb_fences, 2; trained_dont_like_icecream, 2; trained_dont_like_mud, 2; trained_draw_stars_knees, 2; trained_eat_flowers, 2; trained_flap_arms_happy, 2; trained_freckles_feet, 2; trained_hop_puddles, 2; trained_like_sing, 2; trained_scared_ladybugs, 2; trained_sleep_trees, 2; trained_stripes_hair, 2
##
## Dispersion parameter for beta family (): 2.78
##
## Conditional model:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -0.70525 0.12833 -5.495 0.000000039 ***
## num_generics 0.02100 0.01775 1.183 0.2368
## prop_generics 0.35023 0.20626 1.698 0.0895 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Did any particular training features predict prevalence judgments? No
(ps > .19).
## Family: beta ( logit )
## Formula:
## prevalence ~ trained_babies_blankets + trained_bounce_ball_head +
## trained_can_flip_air + trained_chase_shadows + trained_climb_fences +
## trained_dont_like_icecream + trained_dont_like_mud + trained_draw_stars_knees +
## trained_eat_flowers + trained_flap_arms_happy + trained_freckles_feet +
## trained_hop_puddles + trained_like_sing + trained_scared_ladybugs +
## trained_sleep_trees + trained_stripes_hair + (1 | participant) +
## (1 | test_feature)
## Data: data_tidy_dummy
##
## AIC BIC logLik deviance df.resid
## -5546.6 -5406.5 2793.3 -5586.6 8140
##
## Random effects:
##
## Conditional model:
## Groups Name Variance Std.Dev.
## participant (Intercept) 0.6958 0.8341
## test_feature (Intercept) 0.1870 0.4325
## Number of obs: 8160, groups: participant, 507; test_feature, 16
##
## Dispersion parameter for beta family (): 2.77
##
## Conditional model:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -0.550573 0.142345 -3.868 0.00011 ***
## trained_babies_blanketsTRUE 0.034927 0.108064 0.323 0.74654
## trained_bounce_ball_headTRUE -0.047476 0.104438 -0.455 0.64941
## trained_can_flip_airTRUE -0.004133 0.105859 -0.039 0.96885
## trained_chase_shadowsTRUE -0.104644 0.107005 -0.978 0.32810
## trained_climb_fencesTRUE 0.067247 0.103950 0.647 0.51769
## trained_dont_like_icecreamTRUE 0.111003 0.106230 1.045 0.29605
## trained_dont_like_mudTRUE -0.092568 0.109743 -0.844 0.39895
## trained_draw_stars_kneesTRUE 0.101930 0.103258 0.987 0.32357
## trained_eat_flowersTRUE -0.085195 0.103698 -0.822 0.41132
## trained_flap_arms_happyTRUE 0.102891 0.103109 0.998 0.31833
## trained_freckles_feetTRUE -0.028907 0.102005 -0.283 0.77688
## trained_hop_puddlesTRUE 0.062985 0.107716 0.585 0.55873
## trained_like_singTRUE -0.140180 0.107792 -1.300 0.19344
## trained_scared_ladybugsTRUE 0.017752 0.106578 0.167 0.86772
## trained_sleep_treesTRUE 0.003615 0.106654 0.034 0.97296
## trained_stripes_hairTRUE 0.157599 0.105273 1.497 0.13438
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## # A tibble: 16 × 1
## test_feature
## <chr>
## 1 cave
## 2 lion
## 3 potatoes
## 4 banjo
## 5 look_left
## 6 clap
## 7 sad
## 8 maple_syrup
## 9 cats
## 10 opera
## 11 dance
## 12 song
## 13 window
## 14 garbage
## 15 pond
## 16 yellow
Additional possible analyses to do:
Within a condition, find mean prevalence difference for people
who saw feature X vs people who didn’t see the feature –> then
average that, across conditions
feature effects are probably bigger for conditions with fewer
counts
could try to put more limitations on which features are seen in
conditions w fewer counts -
Test features order effects
All participants saw and rated the prevalence of the same set of 16
test features, in random order. Did the order of test feature/prevalence
judgment questions matter for prevalence judgments? No.
## Family: beta ( logit )
## Formula:
## prevalence ~ num_generics * prop_generics * test_feature_order +
## (1 | participant) + (1 | test_feature)
## Data: data_tidy
##
## AIC BIC logLik deviance df.resid
## -5191.6 -5115.3 2606.8 -5213.6 7541
##
## Random effects:
##
## Conditional model:
## Groups Name Variance Std.Dev.
## participant (Intercept) 0.6656 0.8159
## test_feature (Intercept) 0.1945 0.4410
## Number of obs: 7552, groups: participant, 472; test_feature, 16
##
## Dispersion parameter for beta family (): 2.79
##
## Conditional model:
## Estimate Std. Error z value
## (Intercept) -0.763909 0.138140 -5.530
## num_generics 0.030589 0.031044 0.985
## prop_generics 0.573595 0.214778 2.671
## test_feature_order 0.004835 0.004762 1.015
## num_generics:prop_generics -0.017406 0.032531 -0.535
## num_generics:test_feature_order -0.001499 0.001772 -0.846
## prop_generics:test_feature_order -0.016804 0.012242 -1.373
## num_generics:prop_generics:test_feature_order 0.001496 0.001855 0.807
## Pr(>|z|)
## (Intercept) 0.000000032 ***
## num_generics 0.32446
## prop_generics 0.00757 **
## test_feature_order 0.31003
## num_generics:prop_generics 0.59260
## num_generics:test_feature_order 0.39770
## prop_generics:test_feature_order 0.16986
## num_generics:prop_generics:test_feature_order 0.41989
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
