Syntactic Boostrapping MA
Data V2
Anjie Cao and Molly Lewis
2020-08-19
DATA_PATH <- here("data/processed/syntactic_bootstrapping_tidy_data_molly.csv")
ma_data <- read_csv(DATA_PATH) %>%
mutate(row_id = 1:n()) %>%
mutate(agent_argument_type2 = case_when(str_detect(agent_argument_type, "pronoun") ~ "pronoun",
TRUE ~ "noun"),
transitive_event_type2 = case_when(transitive_event_type == "direct_caused_action" ~ "direct_caused_action",
TRUE ~ "indirect_caused_action"))Base model
base_model <- rma.mv(d_calc, d_var_calc, random = ~ 1 | short_cite/same_infant/row_id, data=ma_data)
base_model##
## Multivariate Meta-Analysis Model (k = 53; method: REML)
##
## Variance Components:
##
## estim sqrt nlvls fixed factor
## sigma^2.1 0.0251 0.1584 15 no short_cite
## sigma^2.2 0.1351 0.3675 52 no short_cite/same_infant
## sigma^2.3 0.0000 0.0000 53 no short_cite/same_infant/row_id
##
## Test for Heterogeneity:
## Q(df = 52) = 139.1246, p-val < .0001
##
## Model Results:
##
## estimate se zval pval ci.lb ci.ub
## 0.3250 0.0841 3.8653 0.0001 0.1602 0.4898 ***
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1forest(base_model)
funnel(base_model)
Theoretical moderators
Age
age_model <- rma.mv(d_calc ~ mean_age, d_var_calc, random = ~ 1 | short_cite/same_infant/row_id, data=ma_data)
age_model##
## Multivariate Meta-Analysis Model (k = 53; method: REML)
##
## Variance Components:
##
## estim sqrt nlvls fixed factor
## sigma^2.1 0.0159 0.1263 15 no short_cite
## sigma^2.2 0.1414 0.3760 52 no short_cite/same_infant
## sigma^2.3 0.0000 0.0000 53 no short_cite/same_infant/row_id
##
## Test for Residual Heterogeneity:
## QE(df = 51) = 136.7586, p-val < .0001
##
## Test of Moderators (coefficient 2):
## QM(df = 1) = 1.8208, p-val = 0.1772
##
## Model Results:
##
## estimate se zval pval ci.lb ci.ub
## intrcpt 0.6669 0.2623 2.5430 0.0110 0.1529 1.1810 *
## mean_age -0.0004 0.0003 -1.3494 0.1772 -0.0011 0.0002
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1age_model##
## Multivariate Meta-Analysis Model (k = 53; method: REML)
##
## Variance Components:
##
## estim sqrt nlvls fixed factor
## sigma^2.1 0.0159 0.1263 15 no short_cite
## sigma^2.2 0.1414 0.3760 52 no short_cite/same_infant
## sigma^2.3 0.0000 0.0000 53 no short_cite/same_infant/row_id
##
## Test for Residual Heterogeneity:
## QE(df = 51) = 136.7586, p-val < .0001
##
## Test of Moderators (coefficient 2):
## QM(df = 1) = 1.8208, p-val = 0.1772
##
## Model Results:
##
## estimate se zval pval ci.lb ci.ub
## intrcpt 0.6669 0.2623 2.5430 0.0110 0.1529 1.1810 *
## mean_age -0.0004 0.0003 -1.3494 0.1772 -0.0011 0.0002
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1ggplot(ma_data, aes(x = mean_age, y = d_calc)) +
geom_point() +
geom_smooth(method = "lm")
Vocabulary
vocab_model <- rma.mv(d_calc ~ productive_vocab_median , d_var_calc, random = ~ 1 | short_cite/same_infant/row_id, data=ma_data)
vocab_model##
## Multivariate Meta-Analysis Model (k = 28; method: REML)
##
## Variance Components:
##
## estim sqrt nlvls fixed factor
## sigma^2.1 0.0330 0.1818 5 no short_cite
## sigma^2.2 0.1174 0.3427 28 no short_cite/same_infant
## sigma^2.3 0.1174 0.3427 28 no short_cite/same_infant/row_id
##
## Test for Residual Heterogeneity:
## QE(df = 26) = 71.9795, p-val < .0001
##
## Test of Moderators (coefficient 2):
## QM(df = 1) = 1.1529, p-val = 0.2829
##
## Model Results:
##
## estimate se zval pval ci.lb ci.ub
## intrcpt 0.6330 0.2659 2.3808 0.0173 0.1119 1.1541 *
## productive_vocab_median -0.0054 0.0051 -1.0738 0.2829 -0.0153 0.0045
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1Sentence Structure
structure_model <- rma.mv(d_calc ~ sentence_structure, d_var_calc, random = ~ 1 | short_cite/same_infant/row_id, data=ma_data)
structure_model##
## Multivariate Meta-Analysis Model (k = 53; method: REML)
##
## Variance Components:
##
## estim sqrt nlvls fixed factor
## sigma^2.1 0.0461 0.2148 15 no short_cite
## sigma^2.2 0.0000 0.0001 52 no short_cite/same_infant
## sigma^2.3 0.1137 0.3372 53 no short_cite/same_infant/row_id
##
## Test for Residual Heterogeneity:
## QE(df = 51) = 137.0405, p-val < .0001
##
## Test of Moderators (coefficient 2):
## QM(df = 1) = 3.5176, p-val = 0.0607
##
## Model Results:
##
## estimate se zval pval ci.lb ci.ub
## intrcpt 0.1805 0.1154 1.5637 0.1179 -0.0457 0.4067
## sentence_structuretransitive 0.2475 0.1320 1.8755 0.0607 -0.0111 0.5062
##
## intrcpt
## sentence_structuretransitive .
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1Agent argument type
agent_argument_type_model <- rma.mv(d_calc ~ agent_argument_type2, d_var_calc, random = ~ 1 | short_cite/same_infant/row_id, data=ma_data)
agent_argument_type_model##
## Multivariate Meta-Analysis Model (k = 53; method: REML)
##
## Variance Components:
##
## estim sqrt nlvls fixed factor
## sigma^2.1 0.0078 0.0883 15 no short_cite
## sigma^2.2 0.1385 0.3722 52 no short_cite/same_infant
## sigma^2.3 0.0000 0.0000 53 no short_cite/same_infant/row_id
##
## Test for Residual Heterogeneity:
## QE(df = 51) = 132.1003, p-val < .0001
##
## Test of Moderators (coefficient 2):
## QM(df = 1) = 4.1779, p-val = 0.0410
##
## Model Results:
##
## estimate se zval pval ci.lb ci.ub
## intrcpt 0.1794 0.1058 1.6957 0.0900 -0.0280 0.3868
## agent_argument_type2pronoun 0.3032 0.1483 2.0440 0.0410 0.0125 0.5939
##
## intrcpt .
## agent_argument_type2pronoun *
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1Sink
rma.mv(d_calc ~ agent_argument_type2 +sentence_structure + mean_age, d_var_calc, random = ~ 1 | short_cite/same_infant/row_id, data=ma_data)##
## Multivariate Meta-Analysis Model (k = 53; method: REML)
##
## Variance Components:
##
## estim sqrt nlvls fixed factor
## sigma^2.1 0.0088 0.0936 15 no short_cite
## sigma^2.2 0.0000 0.0001 52 no short_cite/same_infant
## sigma^2.3 0.1302 0.3608 53 no short_cite/same_infant/row_id
##
## Test for Residual Heterogeneity:
## QE(df = 49) = 125.4755, p-val < .0001
##
## Test of Moderators (coefficients 2:4):
## QM(df = 3) = 9.1200, p-val = 0.0277
##
## Model Results:
##
## estimate se zval pval ci.lb
## intrcpt 0.2273 0.3015 0.7540 0.4509 -0.3635
## agent_argument_type2pronoun 0.3324 0.1530 2.1717 0.0299 0.0324
## sentence_structuretransitive 0.2719 0.1372 1.9824 0.0474 0.0031
## mean_age -0.0003 0.0003 -0.8055 0.4205 -0.0009
## ci.ub
## intrcpt 0.8181
## agent_argument_type2pronoun 0.6323 *
## sentence_structuretransitive 0.5408 *
## mean_age 0.0004
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1Transitive event type
rma.mv(d_calc ~ transitive_event_type2, d_var_calc, random = ~ 1 | short_cite/same_infant/row_id, data=ma_data %>% filter(sentence_structure == "transitive"))##
## Multivariate Meta-Analysis Model (k = 26; method: REML)
##
## Variance Components:
##
## estim sqrt nlvls fixed factor
## sigma^2.1 0.2394 0.4893 15 no short_cite
## sigma^2.2 0.0000 0.0000 26 no short_cite/same_infant
## sigma^2.3 0.0000 0.0000 26 no short_cite/same_infant/row_id
##
## Test for Residual Heterogeneity:
## QE(df = 24) = 88.0652, p-val < .0001
##
## Test of Moderators (coefficient 2):
## QM(df = 1) = 1.9307, p-val = 0.1647
##
## Model Results:
##
## estimate se zval pval
## intrcpt 0.4420 0.1511 2.9257 0.0034
## transitive_event_type2indirect_caused_action -0.3969 0.2857 -1.3895 0.1647
## ci.lb ci.ub
## intrcpt 0.1459 0.7381 **
## transitive_event_type2indirect_caused_action -0.9568 0.1630
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1Intransitive event type
rma.mv(d_calc ~ intransitive_event_type , d_var_calc, random = ~ 1 | short_cite/same_infant/row_id, data=ma_data %>% filter(sentence_structure == "intransitive"))##
## Multivariate Meta-Analysis Model (k = 27; method: REML)
##
## Variance Components:
##
## estim sqrt nlvls fixed factor
## sigma^2.1 0.0000 0.0000 11 no short_cite
## sigma^2.2 0.0302 0.1739 27 no short_cite/same_infant
## sigma^2.3 0.0302 0.1739 27 no short_cite/same_infant/row_id
##
## Test for Residual Heterogeneity:
## QE(df = 25) = 46.3175, p-val = 0.0059
##
## Test of Moderators (coefficient 2):
## QM(df = 1) = 1.8094, p-val = 0.1786
##
## Model Results:
##
## estimate se zval pval
## intrcpt 0.0713 0.1438 0.4957 0.6201
## intransitive_event_typeparallel_actions 0.2281 0.1696 1.3452 0.1786
## ci.lb ci.ub
## intrcpt -0.2105 0.3531
## intransitive_event_typeparallel_actions -0.1043 0.5606
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1Methodological moderators
Stimuli Modality
rma.mv(d_calc ~ stimuli_modality, d_var_calc, random = ~ 1 | short_cite/same_infant/row_id, data=ma_data)##
## Multivariate Meta-Analysis Model (k = 53; method: REML)
##
## Variance Components:
##
## estim sqrt nlvls fixed factor
## sigma^2.1 0.0416 0.2040 15 no short_cite
## sigma^2.2 0.1174 0.3427 52 no short_cite/same_infant
## sigma^2.3 0.0000 0.0000 53 no short_cite/same_infant/row_id
##
## Test for Residual Heterogeneity:
## QE(df = 51) = 136.0165, p-val < .0001
##
## Test of Moderators (coefficient 2):
## QM(df = 1) = 1.8477, p-val = 0.1741
##
## Model Results:
##
## estimate se zval pval ci.lb ci.ub
## intrcpt 0.5942 0.2225 2.6704 0.0076 0.1581 1.0304 **
## stimuli_modalityvideo -0.3032 0.2230 -1.3593 0.1741 -0.7403 0.1340
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1Stimuli Actor
rma.mv(d_calc ~ stimuli_actor, d_var_calc, random = ~ 1 | short_cite/same_infant/row_id, data=ma_data)##
## Multivariate Meta-Analysis Model (k = 53; method: REML)
##
## Variance Components:
##
## estim sqrt nlvls fixed factor
## sigma^2.1 0.0394 0.1984 15 no short_cite
## sigma^2.2 0.1176 0.3429 52 no short_cite/same_infant
## sigma^2.3 0.0000 0.0000 53 no short_cite/same_infant/row_id
##
## Test for Residual Heterogeneity:
## QE(df = 51) = 133.7312, p-val < .0001
##
## Test of Moderators (coefficient 2):
## QM(df = 1) = 1.6972, p-val = 0.1927
##
## Model Results:
##
## estimate se zval pval ci.lb ci.ub
## intrcpt 0.4498 0.1346 3.3417 0.0008 0.1860 0.7136 ***
## stimuli_actorperson -0.2073 0.1591 -1.3027 0.1927 -0.5191 0.1046
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1Mass vs distributed
rma.mv(d_calc ~ test_mass_or_distributed, d_var_calc, random = ~ 1 | short_cite/same_infant/row_id, data=ma_data)##
## Multivariate Meta-Analysis Model (k = 53; method: REML)
##
## Variance Components:
##
## estim sqrt nlvls fixed factor
## sigma^2.1 0.0112 0.1058 15 no short_cite
## sigma^2.2 0.1419 0.3767 52 no short_cite/same_infant
## sigma^2.3 0.0000 0.0000 53 no short_cite/same_infant/row_id
##
## Test for Residual Heterogeneity:
## QE(df = 51) = 134.6645, p-val < .0001
##
## Test of Moderators (coefficient 2):
## QM(df = 1) = 3.1204, p-val = 0.0773
##
## Model Results:
##
## estimate se zval pval ci.lb ci.ub
## intrcpt 0.2107 0.1032 2.0411 0.0412 0.0084 0.4131
## test_mass_or_distributedmass 0.2735 0.1548 1.7665 0.0773 -0.0300 0.5770
##
## intrcpt *
## test_mass_or_distributedmass .
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1n_repetitions_sentence
Check this measure
rma.mv(d_calc ~ n_repetitions_sentence, d_var_calc, random = ~ 1 | short_cite/same_infant/row_id, data=ma_data)##
## Multivariate Meta-Analysis Model (k = 53; method: REML)
##
## Variance Components:
##
## estim sqrt nlvls fixed factor
## sigma^2.1 0.0280 0.1673 15 no short_cite
## sigma^2.2 0.1399 0.3740 52 no short_cite/same_infant
## sigma^2.3 0.0000 0.0000 53 no short_cite/same_infant/row_id
##
## Test for Residual Heterogeneity:
## QE(df = 51) = 139.0573, p-val < .0001
##
## Test of Moderators (coefficient 2):
## QM(df = 1) = 0.0589, p-val = 0.8083
##
## Model Results:
##
## estimate se zval pval ci.lb ci.ub
## intrcpt 0.2969 0.1446 2.0532 0.0400 0.0135 0.5803 *
## n_repetitions_sentence 0.0028 0.0114 0.2426 0.8083 -0.0196 0.0251
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1practice_phase
rma.mv(d_calc ~ practice_phase, d_var_calc, random = ~ 1 | short_cite/same_infant/row_id, data=ma_data)##
## Multivariate Meta-Analysis Model (k = 53; method: REML)
##
## Variance Components:
##
## estim sqrt nlvls fixed factor
## sigma^2.1 0.0338 0.1837 15 no short_cite
## sigma^2.2 0.1346 0.3669 52 no short_cite/same_infant
## sigma^2.3 0.0000 0.0000 53 no short_cite/same_infant/row_id
##
## Test for Residual Heterogeneity:
## QE(df = 51) = 138.7878, p-val < .0001
##
## Test of Moderators (coefficient 2):
## QM(df = 1) = 0.5278, p-val = 0.4675
##
## Model Results:
##
## estimate se zval pval ci.lb ci.ub
## intrcpt 0.3801 0.1187 3.2014 0.0014 0.1474 0.6128 **
## practice_phaseyes -0.1084 0.1493 -0.7265 0.4675 -0.4010 0.1841
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1character_identification
rma.mv(d_calc ~ character_identification, d_var_calc, random = ~ 1 | short_cite/same_infant/row_id, data=ma_data)##
## Multivariate Meta-Analysis Model (k = 53; method: REML)
##
## Variance Components:
##
## estim sqrt nlvls fixed factor
## sigma^2.1 0.0372 0.1930 15 no short_cite
## sigma^2.2 0.1315 0.3626 52 no short_cite/same_infant
## sigma^2.3 0.0000 0.0000 53 no short_cite/same_infant/row_id
##
## Test for Residual Heterogeneity:
## QE(df = 51) = 137.6490, p-val < .0001
##
## Test of Moderators (coefficient 2):
## QM(df = 1) = 0.1288, p-val = 0.7197
##
## Model Results:
##
## estimate se zval pval ci.lb ci.ub
## intrcpt 0.2979 0.1102 2.7036 0.0069 0.0819 0.5139
## character_identificationyes 0.0678 0.1889 0.3589 0.7197 -0.3025 0.4381
##
## intrcpt **
## character_identificationyes
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1test_method
rma.mv(d_calc ~ test_method, d_var_calc, random = ~ 1 | short_cite/same_infant/row_id, data=ma_data)##
## Multivariate Meta-Analysis Model (k = 53; method: REML)
##
## Variance Components:
##
## estim sqrt nlvls fixed factor
## sigma^2.1 0.0304 0.1744 15 no short_cite
## sigma^2.2 0.1364 0.3693 52 no short_cite/same_infant
## sigma^2.3 0.0000 0.0000 53 no short_cite/same_infant/row_id
##
## Test for Residual Heterogeneity:
## QE(df = 51) = 137.6641, p-val < .0001
##
## Test of Moderators (coefficient 2):
## QM(df = 1) = 0.2624, p-val = 0.6085
##
## Model Results:
##
## estimate se zval pval ci.lb ci.ub
## intrcpt 0.3125 0.0896 3.4857 0.0005 0.1368 0.4882 ***
## test_methodpoint 0.1884 0.3679 0.5122 0.6085 -0.5326 0.9094
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1Sink - choose moderators with model selcetion method?
rma.mv(d_calc ~ test_mass_or_distributed +
#practice_phase +
character_identification +
test_method +
stimuli_modality,
V = d_var_calc,
random = ~ 1 | short_cite/same_infant/row_id, data = ma_data)##
## Multivariate Meta-Analysis Model (k = 53; method: REML)
##
## Variance Components:
##
## estim sqrt nlvls fixed factor
## sigma^2.1 0.0241 0.1554 15 no short_cite
## sigma^2.2 0.1103 0.3322 52 no short_cite/same_infant
## sigma^2.3 0.0000 0.0000 53 no short_cite/same_infant/row_id
##
## Test for Residual Heterogeneity:
## QE(df = 48) = 115.4569, p-val < .0001
##
## Test of Moderators (coefficients 2:5):
## QM(df = 4) = 8.6070, p-val = 0.0717
##
## Model Results:
##
## estimate se zval pval ci.lb
## intrcpt 0.0804 0.2807 0.2864 0.7746 -0.4698
## test_mass_or_distributedmass 0.5398 0.2231 2.4197 0.0155 0.1026
## character_identificationyes 0.4632 0.2226 2.0811 0.0374 0.0270
## test_methodpoint 0.6020 0.3719 1.6190 0.1054 -0.1268
## stimuli_modalityvideo -0.1824 0.2202 -0.8285 0.4074 -0.6139
## ci.ub
## intrcpt 0.6306
## test_mass_or_distributedmass 0.9771 *
## character_identificationyes 0.8994 *
## test_methodpoint 1.3309
## stimuli_modalityvideo 0.2491
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1mega sink doesn’t add much
rma.mv(d_calc ~ test_mass_or_distributed +
practice_phase +
character_identification +
test_method +
agent_argument_type2 +
sentence_structure +
mean_age,
V = d_var_calc,
random = ~ 1 | short_cite/same_infant/row_id, data = ma_data)##
## Multivariate Meta-Analysis Model (k = 53; method: REML)
##
## Variance Components:
##
## estim sqrt nlvls fixed factor
## sigma^2.1 0.0071 0.0842 15 no short_cite
## sigma^2.2 0.0000 0.0001 52 no short_cite/same_infant
## sigma^2.3 0.1124 0.3353 53 no short_cite/same_infant/row_id
##
## Test for Residual Heterogeneity:
## QE(df = 46) = 106.8245, p-val < .0001
##
## Test of Moderators (coefficients 2:7):
## QM(df = 6) = 15.7473, p-val = 0.0152
##
## Model Results:
##
## estimate se zval pval ci.lb
## intrcpt -0.0866 0.3167 -0.2735 0.7845 -0.7072
## test_mass_or_distributedmass 0.6475 0.1998 3.2413 0.0012 0.2560
## practice_phaseyes 0.0323 0.1499 0.2154 0.8294 -0.2616
## character_identificationyes 0.5098 0.2072 2.4603 0.0139 0.1037
## test_methodpoint 0.6638 0.3494 1.8995 0.0575 -0.0211
## sentence_structuretransitive 0.3021 0.1368 2.2088 0.0272 0.0340
## mean_age -0.0003 0.0003 -0.9613 0.3364 -0.0009
## ci.ub
## intrcpt 0.5340
## test_mass_or_distributedmass 1.0391 **
## practice_phaseyes 0.3262
## character_identificationyes 0.9159 *
## test_methodpoint 1.3487 .
## sentence_structuretransitive 0.5701 *
## mean_age 0.0003
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1