-
Notifications
You must be signed in to change notification settings - Fork 0
/
segundo.R
195 lines (137 loc) · 5.71 KB
/
segundo.R
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
load('/home/amartinezp/Documents/bestWorst/data/01seleccion.RData')
library(R6)
source('/home/amartinezp/lib/load.fenoGAIT2.R')
load.fenoGAIT2()
dat <- merge(dat,df[,'AGE2',drop=FALSE],by='row.names',all.x=TRUE,all.y=FALSE)
row.names(dat) <- dat$Row.names
dat <- dat[,-1]
library(plyr)
library(mvoutlier)
#################################################3
## https://www.analyticsvidhya.com/blog/2016/03/tutorial-powerful-packages-imputing-missing-values/
## https://datascienceplus.com/missing-value-treatment/
## FIRST OF ALL, I RESIDUALIZE THE VARIABLES IN UNIVERSO FOR AGE AND SEX, BECAUSE THEY HAVE A LOT OF EFFECT IN THE FINAL STEPS
covs <- llply(universo.fen, function(X) f1$infoTrait(X)$covariates)
names(covs) <- universo.fen
covs[is.na(covs)] <- 1
formulas <- llply(universo.fen, function(X) paste(X,'~', paste(covs[[X]],collapse='+')))
names(formulas) <- universo.fen
library(solarius)
modelos <- llply(formulas, function(X) solarPolygenic(formula=as.formula(X),data=dat))
kk <- ldply(modelos,function(X) X$resf[,3,drop=FALSE])
kk <- ldply(modelos,function(X) X$resf)
library(reshape)
ll <- cast(kk[,c(1,2,4)],formula= id~ .id )
row.names(ll) <- ll$id
ll <- ll[,-which(colnames(ll)=='id')]
colnames(ll) <- paste0('r',colnames(ll))
###############################
## TEST for imputation ##
###############################
## before imputation one question:
## Can we believe in the imputations?
library(missMDA)
# estimate the number of dimensions to impute
nbdim <- estim_ncpPCA(ll)
ll.df <- as.data.frame(ll)
kk.comp <- MIPCA(ll.df, ncp = max(nbdim$ncp,4), nboot = 1000)
kk.imp <- imputePCA(ll.df, ncp = max(nbdim$ncp,4))
kk.pca <- PCA(kk.imp$completeObs)
plot(kk.comp)
?pool
## Diagnostics
kk.over<-Overimpute(kk.comp)
## little art
library(kandinsky)
kandinsky(kk)
## MIMCA can be used for categorical data
## res <- MIMCA(vnf, ncp=4,nboot=10)
## first impute because pcout not handle missing values
######################
## imputation ##
######################
## 1.- MICE
library(mice)
dis.miss <- c('CVI','VT')
dat <- merge(dat,ll,by='row.names',all.x=TRUE,all.y=FALSE)
row.names(dat) <- dat$id
dat <- dat[,-which(colnames(dat)=='Row.names')]
noResidual <- names(covs)[unlist(lapply(covs,function(X) X[1]==1))]
runiverso.fen <- c(paste0('r',setdiff(universo.fen,noResidual)),noResidual)
kk <- dat[dat$punctured==1,c(runiverso.fen,dis.miss)]
x.mice <- mice(kk,m=10)
kk.mice <- complete(x.mice)
#########
## 2.- Hmisc
library(Hmisc)
imputar <- function(X) {Hmisc::impute(X,fun='random')}
kk.Hmisc <- colwise(imputar)(kk)
row.names(kk.Hmisc) <- row.names(kk)
## i <-39
## i <-38
## table(kk.mice[,i],kk.Hmisc[,i])
## soy conservador y hago la media de ambas imputaciones
kk.imp <- (kk.mice[,runiverso.fen] + kk.Hmisc[,runiverso.fen])/2
kk.imp <- colwise(as.numeric)(kk.imp)
kk.imp$CVI <- factor(kk.Hmisc$CVI)
kk.imp$VT <- factor(kk.Hmisc$VT)
row.names(kk.imp) <- row.names(kk)
source('/home/amartinezp/lib/normalizar.R')
kk.imp.norm <- colwise(normalizar)(kk.imp)
row.names(kk.imp.norm) <- row.names(kk)
colnames(kk.imp.norm) <- paste0('i', colnames(kk))
iruniverso.fen <- paste0('i',runiverso.fen)
dat <- merge(dat,kk.imp.norm,by='row.names',all.x=TRUE,all.y=FALSE)
####################################################
####################################################
## calculo los outliers multivariados
x.out <- pcout(kk.imp.norm[,iruniverso.fen],makeplot=FALSE)
## sum(x.out$wfinal01==0)
sum(x.out$wfinal < 0.05)
## potential location outlier
sum(x.out$wloc < 0.05)
## potential scatter outlier
sum(x.out$wscat < 0.05)
library(MVN)
x.out2 <- mvOutlier(kk.imp.norm[,iruniverso.fen])
mvOutlier(kk.imp.norm[,iruniverso.fen],method='adj.quan',qqplot=TRUE)
## posibles outliers sin residuos 67201,64205
## posibles outliers en los residuos 33301,49403
chungos <- c('67201','64205')
chungos <- c('33301','49403')
chungosMenos <- c('67201','64205')
chungosMenosMenos <- c('53208','62302','67412')
## habria que ver que le pasa a estos individuos
summary(kk.imp.norm[,iruniverso.fen])
kk.imp.norm[chungos,iruniverso.fen]
tr1 <- 'irSTFR';tr2 <- 'irRUSSELL'
tr1 <- 'irHCYhplc'; tr2 <- 'irCYShplc'
tr1 <- 'irSAMhplc'; tr2 <- 'irHCY'
tr1 <- 'irALBUM'; tr2 <- 'irADAMTS13'
tr1 <- 'irFIBc'; tr2 <- 'irFERRIT'
tr1 <- 'irFIXc'; tr2 <- 'irFLAP'
tr1 <- 'irFVIIc'; tr2 <- 'irFIXc'
tr1 <- 'irFVIIc'; tr2 <- 'irFVIIIc'
tr1 <- 'irFvWAg'; tr2 <- 'irFVIIIc'
tr1 <- 'irFvWAg'; tr2 <- 'irPSfunc'
tr1 <- 'irFvWAg'; tr2 <- 'irGGT'
tr1 <- 'irFvWAg'; tr2 <- 'irGAB2'
tr1 <- 'irMAB2'; tr2 <- 'irGAB2'
corr.plot(kk.imp.norm[,tr1],kk.imp.norm[,tr2],xlab=tr1,ylab=tr2)
text(kk.imp.norm[chungos,c(tr1,tr2)],labels=chungos,col='red')
text(kk.imp.norm[chungosMenos,c(tr1,tr2)],labels=chungosMenos,col='green')
dd.plot(kk.imp.norm[,c(tr1,tr2)])
out.multiples <- dd.plot(kk.imp.norm[,iruniverso.fen])
out.multiples$outliers[out.multiples$outliers]
out.multiples$md.rob[out.multiples$md.rob>100]
############################################################################################################################################################
## conclusion: OUTLIERS MULTIVARIADOS
## Hay dos outliers muy jodidos c('67201','64205')
## Hay dos outliers chungos c('53208','62302','67412')
##
## No parecen ser datos erroneos excepto posiblemente para GAB2,MAB2
## NORMALIZO los traits DESPUES de imputar (PRIMERO IMPUTAR)
## La IMPUTACIÓN, la realizo con "mice" y "Hmisc" y luego hago la media entre ambos.
############################################################################################################################################################
dis.miss <- paste0('i',dis.miss)
save(dat,runiverso.fen,iruniverso.fen,universo.dis,universo.extr,universo.fen,universo,chungos,chungosMenos,out.plot,out.dat,dis.miss,file=file.path(out.dat,'dat.kk.imp.nor.RData'))