Moderator: Gordon Smyth

gravatar for Gordon Smyth
Gordon Smyth38k
Reputation:
38,500
Status:
Trusted
Location:
Walter and Eliza Hall Institute of Medical Research, Melbourne, Australia
Website:
http://www.statsci.org...
Scholar ID:
Google Scholar Page
Last seen:
6 hours ago
Joined:
16 years, 8 months ago
Email:
s****@wehi.edu.au

Joint Head of Bioinformatics Division at the Walter and Eliza Hall Institute of Medical Research.

My research group created the limma, edgeR, goseq, Rsubread, csaw and diffHic packages.

Posts by Gordon Smyth

<prev • 3,786 results • page 1 of 379 • next >
1
vote
1
answer
154
views
1
answers
Comment: C: Multiple samples per individual and unbalanced number of samples per individual:
... 1. That's what `duplicateCorrelation` does -- it reduces the influence of subjects with more biopsies vs those with fewer. And more than that, it decides by *how much* to reduce the influence of extra repeated measures. Surely that's what you want. Surely you would want extra biopsies for a subject ...
written 9 days ago by Gordon Smyth38k
2
votes
2
answers
116
views
2
answers
Answer: A: Further clarification on when not to use duplicateCorrelation with technical rep
... > So should duplicateCorrelation be used when mixing biological and technical replicates, but only when there is a minimum number of samples/replicates/levels? Sure, treating factor effects as random often makes more sense when the number of levels is larger, but there is no minimum number. You ...
written 9 days ago by Gordon Smyth38k
1
vote
1
answer
154
views
1
answers
Comment: C: Multiple samples per individual and unbalanced number of samples per individual:
... 1. Looks ok, but you should include Batch only if Batch has a large effect, such that Subjects from the same Batch noticeably cluster together. Otherwise, specifying Subject as a random block would be sufficient. 2. You can only fit correlated blocks in limma because this methodology requires norma ...
written 10 days ago by Gordon Smyth38k
0
votes
3
answers
102
views
3
answers
Answer: A: dream vignette says limma-voom repeated measures two-group comparison needs two
... James Macdonald's comments are on the mark but I think it may be helpful if I give a summary of the situation. The limma Manual and User's Guide do not actually give any explicit advice about how to use `duplicateCorrelation` in conjunction with `voom`, except from carefully documenting the input a ...
written 10 days ago by Gordon Smyth38k
0
votes
2
answers
5.7k
views
2
answers
Comment: C: using duplicateCorrelation with limma+voom for RNA-seq data
... I don't see an advantage in doing that. The weights and the the block correlation only need to be in right ball park. There is no advantage in computing them to 16 significant figures. ...
written 10 days ago by Gordon Smyth38k
1
vote
3
answers
102
views
3
answers
Comment: C: dream vignette says limma-voom repeated measures two-group comparison needs two
... Leandro, it is a good start that you are reading the manual help page for `voom`, but you need to read the "Value" section of the page as well. You can see from this section that `voom()` only returns expression values and weights. It does not return blocks or correlations, so these could not possib ...
written 10 days ago by Gordon Smyth38k
0
votes
1
answer
100
views
1
answers
Comment: C: Setting up contrasts with 'limma', patient data, small number of repeats
... I already gave you the correct design matrix. To test whether the treatments effects are the same for M and F, simply use `contrasts.fit` to compare the two coefficients. Your combination design matrix is over-parametrized and will give uninterpretable results. ...
written 10 days ago by Gordon Smyth38k
3
votes
1
answer
75
views
1
answers
Answer: A: DESeq2 VS edgeR DEG results different in analysis same data
... The DESeq2 statistical tests are analogous to the edgeR glmLRT tests in that they treat the estimated dispersion parameters as if known. Both DESeq2 and edgeR-LRT can give very small p-values and are (in my experience) slightly anti-conservative. The edgeR quasi-likelihood pipeline is intrinsically ...
written 12 days ago by Gordon Smyth38k
0
votes
1
answer
75
views
1
answers
Comment: C: DESeq2 VS edgeR DEG results different in analysis same data
... 1. You say that these are results without filtering, but the edgeR pipelines all assume that you have filtered on expression. You can't omit the filtering step. 2. No one will be able to help you unless provide the code that produced the output you show. ...
written 12 days ago by Gordon Smyth38k
1
vote
1
answer
154
views
1
answers
Answer: A: Multiple samples per individual and different number of samples per individual:
... As you have already noted, this is a straightforward application of a limma two-group comparison with `design <- model.matrix(~Group)` and `block=Subject`. The fact that there is a differing number of samples per subject is not a problem. Note that you can't include Subject in the design matrix - ...
written 12 days ago by Gordon Smyth38k

Latest awards to Gordon Smyth

Scholar 9 weeks ago, created an answer that has been accepted. For A: Limma residuals matrix
Teacher 9 weeks ago, created an answer with at least 3 up-votes. For A: Rsubread, Error in featureCounts paired end
Teacher 9 weeks ago, created an answer with at least 3 up-votes. For A: Limma residuals matrix
Teacher 9 weeks ago, created an answer with at least 3 up-votes. For A: Rsubread, Error in featureCounts paired end
Scholar 9 weeks ago, created an answer that has been accepted. For A: Rsubread, Error in featureCounts paired end
Scholar 10 weeks ago, created an answer that has been accepted. For A: Rsubread, Error in featureCounts paired end
Scholar 12 weeks ago, created an answer that has been accepted. For A: Rsubread, Error in featureCounts paired end
Scholar 12 weeks ago, created an answer that has been accepted. For A: Rsubread, Error in featureCounts paired end
Scholar 12 weeks ago, created an answer that has been accepted. For A: Rsubread, Error in featureCounts paired end
Teacher 12 weeks ago, created an answer with at least 3 up-votes. For A: Rsubread, Error in featureCounts paired end
Scholar 3 months ago, created an answer that has been accepted. For A: Rsubread, Error in featureCounts paired end
Scholar 3 months ago, created an answer that has been accepted. For A: Rsubread, Error in featureCounts paired end
Teacher 3 months ago, created an answer with at least 3 up-votes. For A: Rsubread, Error in featureCounts paired end
Popular Question 9 months ago, created a question with more than 1,000 views. For Columns used by read.maimages for Agilent arrays
Popular Question 9 months ago, created a question with more than 1,000 views. For limma: paired + multiple comparisons + technical replication?
Popular Question 9 months ago, created a question with more than 1,000 views. For non-integer counts for edgeR
Scholar 9 months ago, created an answer that has been accepted. For A: How to add my own Entrez Gene IDs rather than using the ones from a default pack
Scholar 9 months ago, created an answer that has been accepted. For A: barcodeplot how to increase the font size
Popular Question 9 months ago, created a question with more than 1,000 views. For File format for single channel analysis of Agilent microarray data with Limma?
Popular Question 9 months ago, created a question with more than 1,000 views. For total count filter cutoff (edgeR)
Popular Question 9 months ago, created a question with more than 1,000 views. For How do I find up and down regulated genes for each contrast in LIMMA?
Popular Question 9 months ago, created a question with more than 1,000 views. For Estimating fold change from limma 'log2FC' using lumi
Scholar 9 months ago, created an answer that has been accepted. For A: ANOVA-like test via treat() in limma
Popular Question 9 months ago, created a question with more than 1,000 views. For simultaneous use of robust and weighting methods in limma
Scholar 9 months ago, created an answer that has been accepted. For A: How barcode-plot enrichment is calculated?

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 16.09
Traffic: 141 users visited in the last hour