In nim-ach/ASQ3: data from the ASQ-3 study from CADI-UMAG
knitr::opts_chunk$set(collapse = TRUE,
comment = "#>",
dpi = 300)
options(width = 140)
Note: As a priority, this analyses were run using only those with a diagnosis of congenital hypotonia.
Setup details
library(ASQ3)
library(data.table)
dataset <- dataset[diagnostico %like% "hipotonia"]
Working with first assessment
For this page, we'll be fitting different models using generalized additive models (GAM's) using the mgcv package for modeling non-linear relationships, even though it can be used to model linear relationships too. The smoothing parameter (the one that gets to decide how wiggly will be the curve) will be estimated using restricted maximum likelihood (i.e., REML).
About interpretation
The interpretation of the summary of GAM models in R can be seen in this short video (~ 4 minutes) from datacamp.
However, let's get a short intro to a fast interpretation!
Approximate significance of smooth terms:
edf Ref.df F p-value
s(edad_cronologica_meses) 6.475 7.561 28.98 <2e-16 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
This small part of a summary output shows that the influence of edad_cronologica_meses is significant at a p-value of <2e-16. The edf stands for "Effective Degrees of Freedom" and reflects how complex (or wiggly) the relationship is. Thus, an edf of 1 reflects a linear relationship, while and edf of 2 reflects a cuadratic relationship, an edf of 3 cubic relationship and so on.
More information about GAMs can be seen here:
- Generalized Additive Models, Michael Clark: website.
- Nonlinear Models in R: The Wonderful World of mgcv, Noam Ross: youtube.
Modeling
Communication score & chronological age (in months)
Model: comunicacion_total ~ edad_cronologica_meses
plot_gam(data = dataset[n_evaluacion == 1 & edad_cronologica_meses <= 60],
x = edad_cronologica_meses,
y = comunicacion_total,
ylab = "Communication",
xlab = "Months old",
round_to = 5,
method = "REML")
library(mgcv)
mgcv::gam(formula = comunicacion_total ~ s(edad_cronologica_meses),
data = dataset[n_evaluacion == 1 & edad_cronologica_meses <= 60],
method = "REML") |>
summary()
Problem solving score & chronological age (in months)
Model: resolucion_problemas_total ~ edad_cronologica_meses
plot_gam(data = dataset[n_evaluacion == 1 & edad_cronologica_meses <= 60],
x = edad_cronologica_meses,
y = resolucion_problemas_total,
ylab = "Problem solving",
xlab = "Months old",
round_to = 5,
method = "REML")
mgcv::gam(formula = resolucion_problemas_total ~ s(edad_cronologica_meses),
data = dataset[n_evaluacion == 1 & edad_cronologica_meses <= 60],
method = "REML") |>
summary()
Communication score & corrected age (in months)
Model: comunicacion_total ~ edad_corregida_meses
plot_gam(data = dataset[n_evaluacion == 1 & edad_corregida_meses <= 60],
x = edad_corregida_meses,
y = comunicacion_total,
ylab = "Communication",
xlab = "Corrected age (months old)",
round_to = 5,
method = "REML")
mgcv::gam(formula = comunicacion_total ~ s(edad_corregida_meses),
data = dataset[n_evaluacion == 1 & edad_corregida_meses <= 60],
method = "REML") |>
summary()
Fine motor skills score & corrected age (in months)
Model: motora_fina_total ~ edad_corregida_meses
plot_gam(data = dataset[n_evaluacion == 1 & edad_corregida_meses <= 60],
x = edad_corregida_meses,
y = motora_fina_total,
ylab = "Fine motor skills",
xlab = "Corrected age (months old)",
round_to = 5,
method = "REML")
mgcv::gam(formula = motora_fina_total ~ s(edad_corregida_meses),
data = dataset[n_evaluacion == 1 & edad_corregida_meses <= 60],
method = "REML") |>
summary()
Problem solving score & corrected age (in months)
Model: resolucion_problemas_total ~ edad_corregida_meses
plot_gam(data = dataset[n_evaluacion == 1 & edad_corregida_meses <= 60],
x = edad_corregida_meses,
y = resolucion_problemas_total,
ylab = "Fine motor skills",
xlab = "Corrected age (months old)",
round_to = 5,
method = "REML")
mgcv::gam(formula = resolucion_problemas_total ~ s(edad_corregida_meses),
data = dataset[n_evaluacion == 1 & edad_corregida_meses <= 60],
method = "REML") |>
summary()
Socio-individual skills score & corrected age (in months)
Model: socio_individual_total ~ edad_corregida_meses
plot_gam(data = dataset[n_evaluacion == 1 & edad_corregida_meses <= 60],
x = edad_corregida_meses,
y = socio_individual_total,
ylab = "Socio-individual skills",
xlab = "Corrected age (months old)",
round_to = 4,
method = "REML")
mgcv::gam(formula = socio_individual_total ~ s(edad_corregida_meses),
data = dataset[n_evaluacion == 1 & edad_corregida_meses <= 60],
method = "REML") |>
summary()
Communication score & developmental ASQ-3 age (in months)
Model: comunicacion_total ~ asq3_meses
plot_gam(data = dataset[n_evaluacion == 1],
x = asq3_meses,
y = comunicacion_total,
ylab = "Communication",
xlab = "developmental ASQ-3 age (months old)",
round_to = 5,
method = "REML")
mgcv::gam(formula = comunicacion_total ~ s(asq3_meses),
data = dataset[n_evaluacion == 1],
method = "REML") |>
summary()
Problem solving score & developmental ASQ-3 age (in months)
Model: resolucion_problemas_total ~ asq3_meses
plot_gam(data = dataset[n_evaluacion == 1],
x = asq3_meses,
y = resolucion_problemas_total,
ylab = "Problem solving",
xlab = "developmental ASQ-3 age (months old)",
round_to = 5,
method = "REML")
mgcv::gam(formula = resolucion_problemas_total ~ s(asq3_meses),
data = dataset[n_evaluacion == 1],
method = "REML") |>
summary()
Fine motor skills score & developmental ASQ-3 age (in months)
Model: motora_fina_total ~ asq3_meses
plot_gam(data = dataset[n_evaluacion == 1],
x = asq3_meses,
y = motora_fina_total,
ylab = "Fine motor skills",
xlab = "developmental ASQ-3 age (months old)",
round_to = 5,
method = "REML")
mgcv::gam(formula = motora_fina_total ~ s(asq3_meses),
data = dataset[n_evaluacion == 1],
method = "REML") |>
summary()
Fine motor skills score & Communication score
Model: motora_fina_total ~ comunicacion_total
plot_gam(data = dataset[n_evaluacion == 1],
x = comunicacion_total,
y = motora_fina_total,
ylab = "Fine motor skills",
xlab = "Communication",
round_to = 5,
method = "REML")
mgcv::gam(formula = motora_fina_total ~ s(comunicacion_total),
data = dataset[n_evaluacion == 1],
method = "REML") |>
summary()
Socio-individual skills score & Communication score
Model: socio_individual_total ~ comunicacion_total
plot_gam(data = dataset[n_evaluacion == 1],
x = comunicacion_total,
y = socio_individual_total,
ylab = "Socio-individual skills",
xlab = "Communication",
round_to = 5,
method = "REML")
mgcv::gam(formula = socio_individual_total ~ s(comunicacion_total),
data = dataset[n_evaluacion == 1],
method = "REML") |>
summary()
Gross motor skills score score & Communication score
Model: motora_gruesa_total ~ comunicacion_total
plot_gam(data = dataset[n_evaluacion == 1],
x = comunicacion_total,
y = motora_gruesa_total,
ylab = "Gross motor skills",
xlab = "Communication",
round_to = 5,
method = "REML")
mgcv::gam(formula = motora_gruesa_total ~ s(comunicacion_total),
data = dataset[n_evaluacion == 1],
method = "REML") |>
summary()
Looks more like a linear relationship than a non-linear one. This is notorious when we look at the effective degrees of freedom (edf), i.e. the closer to 1 more alike to a linear regression. Although is not significant at a 5% level.
Problem solving score & Fine motor skills score
Model: resolucion_problemas_total ~ motora_fina_total
plot_gam(data = dataset[n_evaluacion == 1],
x = motora_fina_total,
y = resolucion_problemas_total,
ylab = "Problem solving",
xlab = "Fine motor skills",
round_to = 5,
method = "REML")
mgcv::gam(formula = resolucion_problemas_total ~ s(motora_fina_total),
data = dataset[n_evaluacion == 1],
method = "REML") |>
summary()
Socio-individual skills score & Fine motor skills score
Model: socio_individual_total ~ motora_fina_total
plot_gam(data = dataset[n_evaluacion == 1],
x = motora_fina_total,
y = socio_individual_total,
ylab = "Socio-individual skills",
xlab = "Fine motor skills",
round_to = 5,
method = "REML")
mgcv::gam(formula = socio_individual_total ~ s(motora_fina_total),
data = dataset[n_evaluacion == 1],
method = "REML") |>
summary()
Problem solving score & Communication score
Model: resolucion_problemas_total ~ comunicacion_total
plot_gam(data = dataset[n_evaluacion == 1],
x = comunicacion_total,
y = resolucion_problemas_total,
ylab = "Problem solving",
xlab = "Communication",
round_to = 5,
method = "REML")
mgcv::gam(formula = resolucion_problemas_total ~ s(comunicacion_total),
data = dataset[n_evaluacion == 1],
method = "REML") |>
summary()
Problem solving score & Gross motor skills score
Model: resolucion_problemas_total ~ motora_gruesa_total
plot_gam(data = dataset[n_evaluacion == 1],
x = motora_gruesa_total,
y = resolucion_problemas_total,
ylab = "Problem solving",
xlab = "Gross motor skills",
round_to = 5,
method = "REML")
mgcv::gam(formula = resolucion_problemas_total ~ s(motora_gruesa_total),
data = dataset[n_evaluacion == 1],
method = "REML") |>
summary()
Problem solving score & Socio-individual skills score
Model: resolucion_problemas_total ~ socio_individual_total
plot_gam(data = dataset[n_evaluacion == 1],
x = socio_individual_total,
y = resolucion_problemas_total,
ylab = "Problem solving",
xlab = "Socio-individual skills",
round_to = 5,
method = "REML")
mgcv::gam(formula = resolucion_problemas_total ~ s(socio_individual_total),
data = dataset[n_evaluacion == 1],
method = "REML") |>
summary()
Adjusting for evaluation
Communication score & chronological age (in months)
unadjusted <- mgcv::gam(formula = comunicacion_total ~ s(edad_cronologica_meses),
data = dataset[n_evaluacion == 1],
method = "REML")
adjusted <- mgcv::gam(formula = comunicacion_total ~ profesional_id + s(edad_cronologica_meses),
data = dataset[n_evaluacion == 1],
method = "REML")
summary(unadjusted)
summary(adjusted)
anova(unadjusted, adjusted, test = "Chisq")
par(mfrow = c(1,2))
plot(unadjusted, shade = TRUE, xlab = "Edad cronológica", ylab = "Efecto en Habilidades comunicativas", seWithMean = TRUE,
main = "Sin ajuste", ylim = c(-25, 15))
abline(h = c(0,-15))
plot(adjusted, shade = TRUE, xlab = "Edad cronológica", ylab = "Efecto en Habilidades comunicativas", seWithMean = TRUE,
main = "Ajustando por el evaluador", ylim = c(-25, 15))
abline(h = c(0,-15))
Communication score & corrected age (in months)
unadjusted <- mgcv::gam(formula = comunicacion_total ~ s(edad_corregida_meses),
data = dataset[n_evaluacion == 1],
method = "REML")
adjusted <- mgcv::gam(formula = comunicacion_total ~ profesional_id + s(edad_corregida_meses),
data = dataset[n_evaluacion == 1],
method = "REML")
summary(unadjusted)
summary(adjusted)
anova(unadjusted, adjusted, test = "Chisq")
par(mfrow = c(1,3))
plot(unadjusted, shade = TRUE, xlab = "Edad corregida", ylab = "Efecto en Habilidades comunicativas", seWithMean = TRUE,
main = "Sin ajuste", ylim = c(-25, 15))
abline(h = c(0,-15))
plot(adjusted, shade = TRUE, xlab = "Edad corregida", ylab = "Efecto en Habilidades comunicativas", seWithMean = TRUE,
main = "Ajustando por el evaluador", ylim = c(-25, 15))
plot(adjusted, all.terms = TRUE)
nim-ach/ASQ3 documentation built on May 8, 2023, 10:21 p.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
knitr::opts_chunk$set(collapse = TRUE, comment = "#>", dpi = 300) options(width = 140)
Note: As a priority, this analyses were run using only those with a diagnosis of congenital hypotonia.
Setup details
library(ASQ3) library(data.table) dataset <- dataset[diagnostico %like% "hipotonia"]
Working with first assessment
For this page, we'll be fitting different models using generalized additive models (GAM's) using the mgcv package for modeling non-linear relationships, even though it can be used to model linear relationships too. The smoothing parameter (the one that gets to decide how wiggly will be the curve) will be estimated using restricted maximum likelihood (i.e., REML).
About interpretation
The interpretation of the summary of GAM models in R can be seen in this short video (~ 4 minutes) from datacamp. However, let's get a short intro to a fast interpretation!
Approximate significance of smooth terms:
edf Ref.df F p-value
s(edad_cronologica_meses) 6.475 7.561 28.98 <2e-16 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
This small part of a summary output shows that the influence of edad_cronologica_meses is significant at a p-value of <2e-16. The edf stands for "Effective Degrees of Freedom" and reflects how complex (or wiggly) the relationship is. Thus, an edf of 1 reflects a linear relationship, while and edf of 2 reflects a cuadratic relationship, an edf of 3 cubic relationship and so on.
More information about GAMs can be seen here:
- Generalized Additive Models, Michael Clark: website.
- Nonlinear Models in R: The Wonderful World of mgcv, Noam Ross: youtube.
Modeling
Communication score & chronological age (in months)
Model: comunicacion_total ~ edad_cronologica_meses
plot_gam(data = dataset[n_evaluacion == 1 & edad_cronologica_meses <= 60], x = edad_cronologica_meses, y = comunicacion_total, ylab = "Communication", xlab = "Months old", round_to = 5, method = "REML") library(mgcv) mgcv::gam(formula = comunicacion_total ~ s(edad_cronologica_meses), data = dataset[n_evaluacion == 1 & edad_cronologica_meses <= 60], method = "REML") |> summary()
Problem solving score & chronological age (in months)
Model: resolucion_problemas_total ~ edad_cronologica_meses
plot_gam(data = dataset[n_evaluacion == 1 & edad_cronologica_meses <= 60], x = edad_cronologica_meses, y = resolucion_problemas_total, ylab = "Problem solving", xlab = "Months old", round_to = 5, method = "REML") mgcv::gam(formula = resolucion_problemas_total ~ s(edad_cronologica_meses), data = dataset[n_evaluacion == 1 & edad_cronologica_meses <= 60], method = "REML") |> summary()
Communication score & corrected age (in months)
Model: comunicacion_total ~ edad_corregida_meses
plot_gam(data = dataset[n_evaluacion == 1 & edad_corregida_meses <= 60], x = edad_corregida_meses, y = comunicacion_total, ylab = "Communication", xlab = "Corrected age (months old)", round_to = 5, method = "REML") mgcv::gam(formula = comunicacion_total ~ s(edad_corregida_meses), data = dataset[n_evaluacion == 1 & edad_corregida_meses <= 60], method = "REML") |> summary()
Fine motor skills score & corrected age (in months)
Model: motora_fina_total ~ edad_corregida_meses
plot_gam(data = dataset[n_evaluacion == 1 & edad_corregida_meses <= 60], x = edad_corregida_meses, y = motora_fina_total, ylab = "Fine motor skills", xlab = "Corrected age (months old)", round_to = 5, method = "REML") mgcv::gam(formula = motora_fina_total ~ s(edad_corregida_meses), data = dataset[n_evaluacion == 1 & edad_corregida_meses <= 60], method = "REML") |> summary()
Problem solving score & corrected age (in months)
Model: resolucion_problemas_total ~ edad_corregida_meses
plot_gam(data = dataset[n_evaluacion == 1 & edad_corregida_meses <= 60], x = edad_corregida_meses, y = resolucion_problemas_total, ylab = "Fine motor skills", xlab = "Corrected age (months old)", round_to = 5, method = "REML") mgcv::gam(formula = resolucion_problemas_total ~ s(edad_corregida_meses), data = dataset[n_evaluacion == 1 & edad_corregida_meses <= 60], method = "REML") |> summary()
Socio-individual skills score & corrected age (in months)
Model: socio_individual_total ~ edad_corregida_meses
plot_gam(data = dataset[n_evaluacion == 1 & edad_corregida_meses <= 60], x = edad_corregida_meses, y = socio_individual_total, ylab = "Socio-individual skills", xlab = "Corrected age (months old)", round_to = 4, method = "REML") mgcv::gam(formula = socio_individual_total ~ s(edad_corregida_meses), data = dataset[n_evaluacion == 1 & edad_corregida_meses <= 60], method = "REML") |> summary()
Communication score & developmental ASQ-3 age (in months)
Model: comunicacion_total ~ asq3_meses
plot_gam(data = dataset[n_evaluacion == 1], x = asq3_meses, y = comunicacion_total, ylab = "Communication", xlab = "developmental ASQ-3 age (months old)", round_to = 5, method = "REML") mgcv::gam(formula = comunicacion_total ~ s(asq3_meses), data = dataset[n_evaluacion == 1], method = "REML") |> summary()
Problem solving score & developmental ASQ-3 age (in months)
Model: resolucion_problemas_total ~ asq3_meses
plot_gam(data = dataset[n_evaluacion == 1], x = asq3_meses, y = resolucion_problemas_total, ylab = "Problem solving", xlab = "developmental ASQ-3 age (months old)", round_to = 5, method = "REML") mgcv::gam(formula = resolucion_problemas_total ~ s(asq3_meses), data = dataset[n_evaluacion == 1], method = "REML") |> summary()
Fine motor skills score & developmental ASQ-3 age (in months)
Model: motora_fina_total ~ asq3_meses
plot_gam(data = dataset[n_evaluacion == 1], x = asq3_meses, y = motora_fina_total, ylab = "Fine motor skills", xlab = "developmental ASQ-3 age (months old)", round_to = 5, method = "REML") mgcv::gam(formula = motora_fina_total ~ s(asq3_meses), data = dataset[n_evaluacion == 1], method = "REML") |> summary()
Fine motor skills score & Communication score
Model: motora_fina_total ~ comunicacion_total
plot_gam(data = dataset[n_evaluacion == 1], x = comunicacion_total, y = motora_fina_total, ylab = "Fine motor skills", xlab = "Communication", round_to = 5, method = "REML") mgcv::gam(formula = motora_fina_total ~ s(comunicacion_total), data = dataset[n_evaluacion == 1], method = "REML") |> summary()
Socio-individual skills score & Communication score
Model: socio_individual_total ~ comunicacion_total
plot_gam(data = dataset[n_evaluacion == 1], x = comunicacion_total, y = socio_individual_total, ylab = "Socio-individual skills", xlab = "Communication", round_to = 5, method = "REML") mgcv::gam(formula = socio_individual_total ~ s(comunicacion_total), data = dataset[n_evaluacion == 1], method = "REML") |> summary()
Gross motor skills score score & Communication score
Model: motora_gruesa_total ~ comunicacion_total
plot_gam(data = dataset[n_evaluacion == 1], x = comunicacion_total, y = motora_gruesa_total, ylab = "Gross motor skills", xlab = "Communication", round_to = 5, method = "REML") mgcv::gam(formula = motora_gruesa_total ~ s(comunicacion_total), data = dataset[n_evaluacion == 1], method = "REML") |> summary()
Looks more like a linear relationship than a non-linear one. This is notorious when we look at the effective degrees of freedom (edf), i.e. the closer to 1 more alike to a linear regression. Although is not significant at a 5% level.
Problem solving score & Fine motor skills score
Model: resolucion_problemas_total ~ motora_fina_total
plot_gam(data = dataset[n_evaluacion == 1], x = motora_fina_total, y = resolucion_problemas_total, ylab = "Problem solving", xlab = "Fine motor skills", round_to = 5, method = "REML") mgcv::gam(formula = resolucion_problemas_total ~ s(motora_fina_total), data = dataset[n_evaluacion == 1], method = "REML") |> summary()
Socio-individual skills score & Fine motor skills score
Model: socio_individual_total ~ motora_fina_total
plot_gam(data = dataset[n_evaluacion == 1], x = motora_fina_total, y = socio_individual_total, ylab = "Socio-individual skills", xlab = "Fine motor skills", round_to = 5, method = "REML") mgcv::gam(formula = socio_individual_total ~ s(motora_fina_total), data = dataset[n_evaluacion == 1], method = "REML") |> summary()
Problem solving score & Communication score
Model: resolucion_problemas_total ~ comunicacion_total
plot_gam(data = dataset[n_evaluacion == 1], x = comunicacion_total, y = resolucion_problemas_total, ylab = "Problem solving", xlab = "Communication", round_to = 5, method = "REML") mgcv::gam(formula = resolucion_problemas_total ~ s(comunicacion_total), data = dataset[n_evaluacion == 1], method = "REML") |> summary()
Problem solving score & Gross motor skills score
Model: resolucion_problemas_total ~ motora_gruesa_total
plot_gam(data = dataset[n_evaluacion == 1], x = motora_gruesa_total, y = resolucion_problemas_total, ylab = "Problem solving", xlab = "Gross motor skills", round_to = 5, method = "REML") mgcv::gam(formula = resolucion_problemas_total ~ s(motora_gruesa_total), data = dataset[n_evaluacion == 1], method = "REML") |> summary()
Problem solving score & Socio-individual skills score
Model: resolucion_problemas_total ~ socio_individual_total
plot_gam(data = dataset[n_evaluacion == 1], x = socio_individual_total, y = resolucion_problemas_total, ylab = "Problem solving", xlab = "Socio-individual skills", round_to = 5, method = "REML") mgcv::gam(formula = resolucion_problemas_total ~ s(socio_individual_total), data = dataset[n_evaluacion == 1], method = "REML") |> summary()
Adjusting for evaluation
Communication score & chronological age (in months)
unadjusted <- mgcv::gam(formula = comunicacion_total ~ s(edad_cronologica_meses), data = dataset[n_evaluacion == 1], method = "REML") adjusted <- mgcv::gam(formula = comunicacion_total ~ profesional_id + s(edad_cronologica_meses), data = dataset[n_evaluacion == 1], method = "REML") summary(unadjusted) summary(adjusted) anova(unadjusted, adjusted, test = "Chisq") par(mfrow = c(1,2)) plot(unadjusted, shade = TRUE, xlab = "Edad cronológica", ylab = "Efecto en Habilidades comunicativas", seWithMean = TRUE, main = "Sin ajuste", ylim = c(-25, 15)) abline(h = c(0,-15)) plot(adjusted, shade = TRUE, xlab = "Edad cronológica", ylab = "Efecto en Habilidades comunicativas", seWithMean = TRUE, main = "Ajustando por el evaluador", ylim = c(-25, 15)) abline(h = c(0,-15))
Communication score & corrected age (in months)
unadjusted <- mgcv::gam(formula = comunicacion_total ~ s(edad_corregida_meses), data = dataset[n_evaluacion == 1], method = "REML") adjusted <- mgcv::gam(formula = comunicacion_total ~ profesional_id + s(edad_corregida_meses), data = dataset[n_evaluacion == 1], method = "REML") summary(unadjusted) summary(adjusted) anova(unadjusted, adjusted, test = "Chisq") par(mfrow = c(1,3)) plot(unadjusted, shade = TRUE, xlab = "Edad corregida", ylab = "Efecto en Habilidades comunicativas", seWithMean = TRUE, main = "Sin ajuste", ylim = c(-25, 15)) abline(h = c(0,-15)) plot(adjusted, shade = TRUE, xlab = "Edad corregida", ylab = "Efecto en Habilidades comunicativas", seWithMean = TRUE, main = "Ajustando por el evaluador", ylim = c(-25, 15)) plot(adjusted, all.terms = TRUE)
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.