CODEWITHSUNDEEP
raggregatemean
psoares
psoares

Reputation: 4883

merge list of lists into one by averaging each value in lists r

So I'm not really sure about the best way to achieve what I want to do.

Here is my problem: I'm using multiple imputation in R with the package mice and I'm using the function cox.zph to get the residuals for each imputed model. I would like to average all the imputed residuals to get only one value for each variable. And honestly I have no idea how to achieve this. I've looked at lapply and mapply. 

res.scho <- NULL
for (i in 1:m) {
    res.scho[[i]] <- cox.zph(modf$analyses[[i]])
}

res.scho[[1]]
              rho  chisq        p
Alc=1       0.0622  0.552 4.58e-01
HIV=1       0.1050  1.873 1.71e-01
Diabetes=1 -0.1240  2.227 1.36e-01
age        -0.1388  2.877 8.99e-02
GLOBAL          NA 44.467 5.97e-08

Basically I have m imputed dataset and res.scho[[m]] and I would like to combine each res.scho into one.

I'm still very stuck at forloops that I used in other languages and I'm having some problems to use mapply. This could be one of the issue. Nevertheless, I would be really grateful if someone could give me some pointers to help me achieve this and better use R.

Thank you!

EDIT

EXpected output:

Let's say I have two imputed datasets. m=2

res.scho[[1]]
              rho  chisq        p
Alc=1       0.0622  0.552 4.58e-01
HIV=1       0.1050  1.873 1.71e-01
Diabetes=1 -0.1240  2.227 1.36e-01
age        -0.1388  2.877 8.99e-02
GLOBAL          NA 44.467 5.97e-08

res.scho[[2]]
              rho  chisq        p
Alc=1       0.0522  0.752 5.58e-01
HIV=1       0.1550  1.473 2.71e-01
Diabetes=1 -0.1140  2.927 4.36e-01
age        -0.1188  2.077 3.99e-02
GLOBAL          NA 44.400 7.97e-08

My desired output would be in the same form as the list but the average of the two for each value, for instance:

 Average_res.scho
              rho  chisq        p
Alc=1       0.0572  0.652 5.08e-01
HIV=1       0.1300  1.673 2.21e-01
Diabetes=1 -0.1190  2.577 2.86e-01
age        -0.1288  2.477 4.24e-02
GLOBAL          NA 44.433 6.97e-08

for instance the column of rho is obtained by colum (res.scho[[1]] + res.scho[[2]]) /2

EDIT1

Following the suggestions of konvas I'm trying to use his ideas to get the desired output. Here is what I have so far:

rho <- NULL
chisq <- NULL
p <- NULL
for (i in 1:70) {
    rho[[i]] <- res.scho[[i]]$table[,"rho"]
    chisq[[i]] <- res.scho[[i]]$table[,"chisq"]
    p[[i]] <- res.scho[[i]]$table[,"p"]

I've extracted a list of lists for each column of res.scho - not the perfect solution. So if I do rho[[1]] I see the first column

 [[1]]
  CliForm=1       Sit=2       Alc=1       HIV=1  Diabetes=1         age      GLOBAL 
  0.17300198 -0.45800541  0.06224951  0.10495093 -0.12401631 -0.13879592          NA

Now, I will thinking about to this for rho, chisq and p:

for (i in 1:70) {
    result <- sapply(names(rho[[1]]),
             function(x) colMeans(sapply(rho, "[[", x)))
}

And I get the following error that I've been trying to solve:

 Error in colMeans(sapply(rho, "[[", x)) :   'x' must be an array of at least two dimensions

Upvotes: 0

Views: 217

Answers (1)

konvas
konvas

Reputation: 14346

First, extract the relevant matrix of coefficients for each entry of the list res.scho:

res.scho.tables <- lapply(res.scho, `[[`, "table")

Then since all you want is an average (and matrices can be added up elementwise quite fast) you can call 

Average_res.scho <- do.call(`+`, res.scho.tables) / length(res.scho.tables)

Upvotes: 1

Related Questions