neuroimagen:adni_data
Table of Contents
ADNI data R package
Los datos de ADNI estan empaquetados en formato R por Michel Donahue. El paquete se llama ADNIMERGE y está instalado en el servidor. Puede descargarse de la pagina de ADNI. Hay además un grupo de google de adnimerge que puede ser util.
Instalación
Copiando del README:
ADNI Data Package for R
For users of R, we have developed a data package "ADNIMERGE" which contains coded data, documentation, and analysis vignettes. It depends on Frank Harrrel's Hmisc package which can be installed from the R package repository (CRAN) by:
R prompt> install.packages("Hmisc")
After downloading the compressed ADNIMERGE_0.0.1.tar.gz file, it can be installed to your R system by entering the following at an R prompt:
R prompt> install.packages("/path/to/file/ADNIMERGE_0.0.1.tar.gz", repos = NULL, type = "source")
If you have trouble installing, more information can be found at http://cran.r-project.org/doc/manuals/R-admin.html#Installing-packages.
To browse document and analysis vignettes,
R prompt> help(package = "ADNIMERGE")
To view, for example, documentation for the adas:
R prompt> ?adas
Cargar la biblioteca
library("ADNIMERGE")
MCI due to AD
Los MCI por AD se sacan:
dxmci <- dxsum[dxsum[, "DXMDUE"] == "MCI due to Alzheimer's Disease" & !is.na(dxsum$DXMDUE),]
y además amnesicos
dxmci=dxsum[dxsum[, "DXMDUE"] == "MCI due to Alzheimer's Disease" & !is.na(dxsum$DXMDUE) & dxsum[, "DXMDES"] == "MCI - Memory features (amnestic)" & !is.na(dxsum$DXMDES),]
First try
mcimerged0 <- merge(dxmci, adnimerge, by=c("RID", "VISCODE"))
mciwav <- mcimerged0[!is.na(mcimerged0$AV45) ,]
ahora voy a añadir las variables de delay real que me ha calculado sergi,
comp_delay_recall <- read.csv("adni_delay_recall.csv", sep=",")
mci_dr <- merge(adnimerge, comp_delay_recall, by=c("RID", "VISCODE"))
Merging from several tables
Merging and calculating data
mt1 <- merge(adas, adnimerge, by=c("RID", "VISCODE") )
mt1$vAGE = mt1$AGE + mt1$Years
a <-lm(mt1$Hippocampus ~ mt1$ICV)
b=a$coefficients[[2]]
mt1$aHV = mt1$Hippocampus - b*(mt1$ICV - mean(mt1$ICV, na.rm = TRUE))
Preparing data to fit
data <- data.frame(mt1$aHV, mt1$vAGE, mt1$PTGENDER, mt1$PTEDUCAT, mt1$Q4SCORE) datac <- data[complete.cases(data),]
Just fitting
fit <- cusp(y ~ mt1.Q4SCORE, alpha ~ mt1.aHV +mt1.vAGE + mt1.PTGENDER +mt1.PTEDUCAT, beta ~ mt1.aHV +mt1.vAGE + mt1.PTGENDER +mt1.PTEDUCAT, datac)
A working example, for FDG data and NEUROBAT ADNI table
mt1 <- merge(neurobat, adnimerge, by=c("RID", "VISCODE") )
mt1$vAGE = mt1$AGE + mt1$Years
data <- data.frame(mt1$FDG, mt1$vAGE, mt1$PTGENDER, mt1$PTEDUCAT, mt1$AVDEL30MIN)
datac <- data[complete.cases(data),]
fit <- cusp(y ~ mt1.AVDEL30MIN, alpha ~ mt1.FDG +mt1.vAGE + mt1.PTGENDER +mt1.PTEDUCAT, beta ~ mt1.FDG +mt1.vAGE + mt1.PTGENDER +mt1.PTEDUCAT, datac)
y lo miramos a ver que tal queda:
> summary(fit)
Call:
cusp(formula = y ~ mt1.AVDEL30MIN, alpha = alpha ~ mt1.FDG +
mt1.vAGE + mt1.PTGENDER + mt1.PTEDUCAT, beta = beta ~ mt1.FDG +
mt1.vAGE + mt1.PTGENDER + mt1.PTEDUCAT, data = datac)
Deviance Residuals:
Min 1Q Median 3Q Max
-3.087409 -0.279966 -0.002222 0.592400 3.437777
Coefficients:
Estimate Std. Error z value Pr(>|z|)
a[(Intercept)] -3.481541 0.293956 -11.844 < 2e-16 ***
a[mt1.FDG] 0.418469 0.025365 16.498 < 2e-16 ***
a[mt1.vAGE] -0.010894 0.002552 -4.269 1.96e-05 ***
a[mt1.PTGENDERFemale] 0.282399 0.036690 7.697 1.39e-14 ***
a[mt1.PTEDUCAT] 0.069970 0.007042 9.936 < 2e-16 ***
b[(Intercept)] 5.548930 0.399315 13.896 < 2e-16 ***
b[mt1.FDG] -0.555766 0.004995 -111.262 < 2e-16 ***
b[mt1.vAGE] -0.011760 0.004163 -2.825 0.00473 **
b[mt1.PTGENDERFemale] 0.308486 0.061716 4.998 5.78e-07 ***
b[mt1.PTEDUCAT] 0.027550 0.011087 2.485 0.01296 *
w[(Intercept)] -1.827114 0.016373 -111.594 < 2e-16 ***
w[mt1.AVDEL30MIN] 0.262274 0.002336 112.278 < 2e-16 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Null deviance: 4464.2 on 3319 degrees of freedom
Linear deviance: 50944.0 on 3314 degrees of freedom
Logist deviance: NA on NA degrees of freedom
Delay deviance: 1690.2 on 3308 degrees of freedom
R.Squared logLik npar AIC AICc BIC
Linear model 0.2150188 -9243.943 6 18499.885 18499.910 18536.531
Cusp model 0.6407748 -3427.447 12 6878.895 6878.989 6952.188
---
Note: R.Squared for cusp model is Cobb's pseudo-R^2. This value
can become negative.
Chi-square test of linear vs. cusp model
X-squared = 1.163e+04, df = 6, p-value = 0
Number of optimization iterations: 82
cusp3d(fit) give the 3D graph,
plot(fit) give the proyection on control plane and more info
How to do this with Composite Scores
tmp_np <- merge(adas, neurobat, by=c("RID", "VISCODE") )
mt2fa <- merge(tmp_np, adnimerge, by=c("RID", "VISCODE") )
rm(tmp_np)
mt2fa$vAGE = mt2fa$AGE + mt2fa$Years
data <- data.frame(mt2fa$FDG, mt2fa$vAGE, mt2fa$PTGENDER, mt2fa$PTEDUCAT, mt2fa$Q4SCORE, mt2fa$AVDEL30MIN)
datac <- data[complete.cases(data),]
datac$zDR = (datac$mt2fa.Q4SCORE - mean(datac$mt2fa.Q4SCORE))/sd(datac$mt2fa.Q4SCORE)
datac$zAVD = (mean(datac$mt2fa.AVDEL30MIN) - datac$mt2fa.AVDEL30MIN)/sd(datac$mt2fa.AVDEL30MIN)
gfam <- data.frame(datac$zDR, datac$zAVD)
famod <- fa(gfam, scores="regression")
datac$DRCS <- famod$scores
fitdr <- cusp(y ~ DRCS, alpha ~ mt2fa.FDG +mt2fa.vAGE + mt2fa.PTGENDER +mt2fa.PTEDUCAT, beta ~ mt2fa.FDG +mt2fa.vAGE + mt2fa.PTGENDER +mt2fa.PTEDUCAT, datac)
The composites can be seen here:
pairs.panels(datac, pch=21)
Now, notice that your results will be as good as your former worsest (or even worse!).
> summary(fitdr)
Call:
cusp(formula = y ~ DRCS, alpha = alpha ~ mt2fa.FDG + mt2fa.vAGE +
mt2fa.PTGENDER + mt2fa.PTEDUCAT, beta = beta ~ mt2fa.FDG +
mt2fa.vAGE + mt2fa.PTGENDER + mt2fa.PTEDUCAT, data = datac)
Deviance Residuals:
Min 1Q Median 3Q Max
-3.60924 -0.63803 -0.05687 0.35020 3.02185
Coefficients:
Estimate Std. Error z value Pr(>|z|)
a[(Intercept)] 4.861677 0.302969 16.047 < 2e-16 ***
a[mt2fa.FDG] -0.658838 0.018241 -36.119 < 2e-16 ***
a[mt2fa.vAGE] 0.010701 0.002784 3.844 0.000121 ***
a[mt2fa.PTGENDERFemale] -0.266786 0.007268 -36.705 < 2e-16 ***
a[mt2fa.PTEDUCAT] -0.073632 0.007562 -9.737 < 2e-16 ***
b[(Intercept)] 5.945053 0.365065 16.285 < 2e-16 ***
b[mt2fa.FDG] -0.741062 NA NA NA
b[mt2fa.vAGE] -0.017238 0.004638 -3.717 0.000202 ***
b[mt2fa.PTGENDERFemale] 0.600601 0.040893 14.687 < 2e-16 ***
b[mt2fa.PTEDUCAT] 0.042803 0.012205 3.507 0.000453 ***
w[(Intercept)] 0.426748 0.014152 30.154 < 2e-16 ***
w[DRCS] 1.169788 0.008142 143.666 < 2e-16 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Null deviance: 3895.6 on 3317 degrees of freedom
Linear deviance: 2026.9 on 3312 degrees of freedom
Logist deviance: NA on NA degrees of freedom
Delay deviance: 2475.7 on 3306 degrees of freedom
R.Squared logLik npar AIC AICc BIC
Linear model 0.2880251 -3890.378 6 7792.755 7792.781 7829.398
Cusp model 0.3869597 -3622.986 12 7269.971 7270.066 7343.257
---
Note: R.Squared for cusp model is Cobb's pseudo-R^2. This value
can become negative.
Chi-square test of linear vs. cusp model
X-squared = 534.8, df = 6, p-value = 0
Number of optimization iterations: 43
neuroimagen/adni_data.txt · Last modified: 2020/08/04 10:58 by 127.0.0.1