Open miguecass opened 8 months ago
Hi Miguel,
Thank you for your questions.
To get the covariance matrix, you could run the following as illustrated in the tutorial
cov= matrix(NA,12,12) # 12 is the number of the coefs in your design matrix
gene_i = 1 # gene_i is the index of the gene of which you want the covariance
cov[lower.tri(cov,diag=T)] = as.numeric(re$covariance[gene_i,])
cov[upper.tri(cov)] = t(cov)[upper.tri(cov)] # cov is the covariance
Best regards,
Liang
On Fri, Oct 13, 2023 at 7:34 AM miguecass @.***> wrote:
Hi,
I ran your method for my data and it was surprisingly fast, so congrats on that first. However, I have some doubts:
- In the output random effect is always NULL in my case. I assumed the random effect is the second input in the function nebula? Otherwise I am not sure where to put it, because in your tutorial it was not included in the design.
re = nebula(data,metadata$Batch,pred=df, covariance = TRUE)
- Another doubt is about the covariances. I have 78 covariances but they are not named. So I am not sure about the order of the namings. My design matrix has 12 coefficients but I would like to know what is the order of the coefficients because I need to compute some complex contrasts.
Thanks in advance,
Best Miguel
— Reply to this email directly, view it on GitHub https://github.com/lhe17/nebula/issues/37, or unsubscribe https://github.com/notifications/unsubscribe-auth/AGDISURJUFTTCGJ7DFQZ2P3X7ERK5ANCNFSM6AAAAAA57ANPHU . You are receiving this because you are subscribed to this thread.Message ID: @.***>
Thanks for the response. Even if output_re is FALSE the fixed coefficients from the model are already taking into account the random effect, right?
Hi Miguel,
Yes, that is correct.
Best regards, Liang
On Sat, Oct 14, 2023 at 11:23 AM miguecass @.***> wrote:
Thanks for the response. Even if output_re is FALSE the fixed coefficients from the model are already taking into account the random effect, right?
— Reply to this email directly, view it on GitHub https://github.com/lhe17/nebula/issues/37#issuecomment-1762965594, or unsubscribe https://github.com/notifications/unsubscribe-auth/AGDISUQAGQDGRO5WA4U5QRLX7KU6BANCNFSM6AAAAAA57ANPHU . You are receiving this because you commented.Message ID: @.***>
Hi,
I ran your method for my data and it was surprisingly fast, so congrats on that first. However, I have some doubts:
1) In the output random effect is always NULL in my case. I assumed the random effect is the second input (
metadata$Batch
) in the function nebula? Otherwise I am not sure where to put it, because in your tutorial it was not included in the design.re = nebula(data,metadata$Batch,pred=df, covariance = TRUE)
2) Another doubt is about the covariances. I have 78 covariances but they are not named. So I am not sure about the order of the namings. My design matrix has 12 coefficients but I would like to know what is the order of the covariances since I have 78 of them and I need to compute some complex contrasts.
EDIT: Actually thinking it a bit more, would I need the covariance matrix to compute DE genes for my contrasts? Can I just not multiply the coefficients for each gene by my contrast matrix? (Same as with limma approach)
Thanks in advance,
Best Miguel