I have a bunch of parameters to adjust for continuous/categorical covariates (age and sex) and I found out about using residuals as adjusted data. I have one question about this procedure. We possibly get negatives and positives given that residual= original-predicted. Do we need to take only the absolute value of the residuals as the adjusted data? any help on this is kindly appreciated.
Here is a summary of my question.
Let's say I have parameter 'Z' corresponding ages are 'age'.
Z <- c(0.9,1.2,3.0,4.5,0.8,0.4)
age <- c(30,22,45,60,33,20)
fit <-lm(Z~age)
adjusted<- fit$residuals + fit$coefficients["(Intercept)"] # this is the residual
This gives -2.1 -1.0 -1.5 -1.5 -2.5 -1.6 as adjusted Z for age bias. but Z supposed to be positives. Is this acceptable. Can I report these negatives as adjusted values for covariate age or should I get the absolute value of residuals?
A similar post is here: Adjust for covariates
0 answers
No answers yet.
Log in to answer this question.
-
ANOVA or linear regression
written by Penny •
-
LInear Models with HLA-B alleles in R
written by schagas •
-
Design matrix for microarray experiment with covariates, cofactors, multiple factors, and one trea…
written by bkleiboeker •
-
Plotting residuals for limma lmfit model
written by ritz •
-
Covariate correction which data take for downstream analysis?
written by camillab. •
-
Batch effect removal with the help of sva
written by Gene_MMP8
-
how to adjust a phenotype according to age
written by xinhui.wang
-
Finding the best model for expression data (RNA-Seq counts)
written by User 7754
-
how to create lower or upper triangle distance matrix in R?
written by blacktomato27
-
Volcano Plot from DEseq2
written by 1769mkc
You should include some example data and code, otherwise it would be difficult to give any sort of specific example.
You can get residual with
adjusted <- resid(fit)And it is ok with negatives, as Z is now a different variable. It is the residual of Z on age and it doesn't need to fall within a predefined range.