Complexity and Roy et al. (2015) Wordbirth entropy analyses
Analysis: Is there a relationship between the Roy et al. (2015) wordbirth measures of entropy and our measures of complexity?
Three measures of entropy – temporal, spatial, and linguistic.
Read in data, etc.
roy = read.csv("roy_et_al_data.csv")
eng.complexity = read.csv("complexityNormsEnglishexp9.csv")
d = roy %>%
inner_join(eng.complexity)
d$word = as.factor(d$word)
d$mean.KL = rowMeans(cbind(d$srl.topic.KL, d$srl.temp.KL, d$srl.sp.KL))m = cor(d[,c(-1,-3:-4,-14, -17:-18)], use = "complete")
corrplot(m)
ggplot(d, aes(x= complexity, y= srl.sp.KL, label=word)) +
geom_text(aes(label=word),hjust=0, vjust=0) +
geom_smooth(method = "lm") +
facet_grid(.~ small.cat) +
themeML 
ggplot(d, aes(x= complexity, y= srl.topic.KL, label=word)) +
geom_text(aes(label=word),hjust=0, vjust=0) +
geom_smooth(method = "lm") +
facet_grid(.~ small.cat) +
themeML 
ggplot(d, aes(x= complexity, y= srl.temp.KL, label=word)) +
geom_text(aes(label=word),hjust=0, vjust=0) +
geom_smooth(method = "lm") +
facet_grid(.~ small.cat) +
themeML 
ggplot(d, aes(x= complexity, y= mean.KL, label=word)) +
geom_text(aes(label=word),hjust=0, vjust=0) +
geom_smooth(method = "lm") +
facet_grid(.~ small.cat) +
themeML 
ggplot(d, aes(x= complexity, y= aoa, label=word)) +
geom_text(aes(label=word),hjust=0, vjust=0) +
geom_smooth(method = "lm") +
facet_grid(.~ small.cat) +
themeML 
ggplot(d, aes(x= complexity, y= sln.freq.pre, label=word)) +
geom_text(aes(label=word),hjust=0, vjust=0) +
geom_smooth(method = "lm") +
facet_grid(.~ small.cat) +
themeML 
ggplot(d, aes(x= complexity, y= s.uttlen.pre, label=word)) +
geom_text(aes(label=word),hjust=0, vjust=0) +
geom_smooth(method = "lm") +
facet_grid(.~ small.cat) +
themeML 
Some models
summary(lmer(complexity ~ srl.sp.KL +s.cmu.phon + sln.freq.pre + s.uttlen.pre + (1|cdi.cat), d))## Linear mixed model fit by REML ['lmerMod']
## Formula:
## complexity ~ srl.sp.KL + s.cmu.phon + sln.freq.pre + s.uttlen.pre +
## (1 | cdi.cat)
## Data: d
##
## REML criterion at convergence: 218.4
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -1.75133 -0.66804 -0.01131 0.66170 2.48054
##
## Random effects:
## Groups Name Variance Std.Dev.
## cdi.cat (Intercept) 0.1705 0.4130
## Residual 0.4911 0.7008
## Number of obs: 91, groups: cdi.cat, 19
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 2.71988 0.14618 18.607
## srl.sp.KL -0.12239 0.07742 -1.581
## s.cmu.phon 0.48049 0.14743 3.259
## sln.freq.pre 0.11012 0.11436 0.963
## s.uttlen.pre -0.07002 0.12527 -0.559
##
## Correlation of Fixed Effects:
## (Intr) sr..KL s.cm.p sln.f.
## srl.sp.KL -0.013
## s.cmu.phon 0.339 -0.005
## sln.freq.pr -0.026 -0.241 0.450
## s.uttlen.pr -0.248 0.410 -0.143 -0.395summary(lmer(complexity ~ srl.temp.KL +s.cmu.phon + sln.freq.pre + s.uttlen.pre + (1|cdi.cat), d))## Linear mixed model fit by REML ['lmerMod']
## Formula:
## complexity ~ srl.temp.KL + s.cmu.phon + sln.freq.pre + s.uttlen.pre +
## (1 | cdi.cat)
## Data: d
##
## REML criterion at convergence: 219
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -1.72558 -0.69846 -0.05592 0.61388 2.58257
##
## Random effects:
## Groups Name Variance Std.Dev.
## cdi.cat (Intercept) 0.1517 0.3895
## Residual 0.5037 0.7097
## Number of obs: 91, groups: cdi.cat, 19
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 2.71235 0.14330 18.928
## srl.temp.KL -0.10831 0.08568 -1.264
## s.cmu.phon 0.49020 0.14911 3.288
## sln.freq.pre 0.07396 0.11209 0.660
## s.uttlen.pre -0.02517 0.11822 -0.213
##
## Correlation of Fixed Effects:
## (Intr) sr..KL s.cm.p sln.f.
## srl.temp.KL 0.012
## s.cmu.phon 0.345 -0.081
## sln.freq.pr -0.031 -0.064 0.472
## s.uttlen.pr -0.259 0.237 -0.167 -0.345summary(lm(complexity ~ srl.topic.KL +s.cmu.phon + sln.freq.pre + s.uttlen.pre, d))##
## Call:
## lm(formula = complexity ~ srl.topic.KL + s.cmu.phon + sln.freq.pre +
## s.uttlen.pre, data = d)
##
## Residuals:
## Min 1Q Median 3Q Max
## -1.39258 -0.56063 -0.04917 0.57341 2.40378
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 2.69086 0.10995 24.473 <2e-16 ***
## srl.topic.KL 0.05775 0.09699 0.595 0.553
## s.cmu.phon 0.38858 0.16228 2.395 0.019 *
## sln.freq.pre 0.07729 0.12463 0.620 0.537
## s.uttlen.pre 0.02246 0.11528 0.195 0.846
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.782 on 80 degrees of freedom
## (6 observations deleted due to missingness)
## Multiple R-squared: 0.08181, Adjusted R-squared: 0.0359
## F-statistic: 1.782 on 4 and 80 DF, p-value: 0.1406summary(lm(complexity ~ srl.temp.KL +s.cmu.phon + sln.freq.pre + s.uttlen.pre, d))##
## Call:
## lm(formula = complexity ~ srl.temp.KL + s.cmu.phon + sln.freq.pre +
## s.uttlen.pre, data = d)
##
## Residuals:
## Min 1Q Median 3Q Max
## -1.5689 -0.5431 -0.1201 0.4984 2.4805
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 2.68043 0.10845 24.715 < 2e-16 ***
## srl.temp.KL -0.12425 0.08810 -1.410 0.16204
## s.cmu.phon 0.44102 0.15410 2.862 0.00529 **
## sln.freq.pre 0.04555 0.11494 0.396 0.69288
## s.uttlen.pre -0.03603 0.11464 -0.314 0.75408
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.7759 on 86 degrees of freedom
## Multiple R-squared: 0.1175, Adjusted R-squared: 0.0765
## F-statistic: 2.864 on 4 and 86 DF, p-value: 0.02798summary(lm(complexity ~ srl.sp.KL +s.cmu.phon + sln.freq.pre + s.uttlen.pre, d))##
## Call:
## lm(formula = complexity ~ srl.sp.KL + s.cmu.phon + sln.freq.pre +
## s.uttlen.pre, data = d)
##
## Residuals:
## Min 1Q Median 3Q Max
## -1.43541 -0.48384 -0.03978 0.54140 2.29662
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 2.67962 0.10848 24.702 < 2e-16 ***
## srl.sp.KL -0.11069 0.07957 -1.391 0.16776
## s.cmu.phon 0.42633 0.15359 2.776 0.00676 **
## sln.freq.pre 0.06842 0.11636 0.588 0.55810
## s.uttlen.pre -0.06104 0.12071 -0.506 0.61438
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.7762 on 86 degrees of freedom
## Multiple R-squared: 0.117, Adjusted R-squared: 0.07594
## F-statistic: 2.849 on 4 and 86 DF, p-value: 0.02861cor.test(d[d$small.cat == "Nouns", "complexity"],
d[d$small.cat == "Nouns", "srl.topic.KL"])##
## Pearson's product-moment correlation
##
## data: d[d$small.cat == "Nouns", "complexity"] and d[d$small.cat == "Nouns", "srl.topic.KL"]
## t = 1.9511, df = 34, p-value = 0.05932
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.01252098 0.58488136
## sample estimates:
## cor
## 0.3173201None of the entropy measures are reliable predicotrs of complexity, even for controling for stuff. There’s a marginal relationship betwen srl.topic.KL and complexity, for nouns only (n = 36, p=.06, r = .32).