Stitistical significance consultation
Hi,
If somebody comes to you for getting consult on statistical significance in his/her gene expression what would be your suggestion? I will ask about FDR threshold and how stringent should be in extracting the top genes. Anything else that I missed please????
Thank you
0 answers
No answers yet.
Log in to answer this question.
More posts like this
-
Which Should Take Priority When Selecting DEGs: FDR or Log Fold Change?
written by bioinfo1994 •
-
Relationship between FDR alpha level and adjusted p-value thresholds
written by basch •
-
Will it be right to select a random threshold in GWAS to declare significant peaks?
written by biotechnology415 •
-
Getting RNA expression from TCGA only for these patients IDs
written by zizigolu
-
PCA-based clustering gene expression data
written by Gene_MMP8
-
Differential gene expression using GWAS/SNPs data
written by Star •
-
If I am right about fold change in RNA-seq
written by zizigolu
-
DESeq2: Find differentially expressed genes specific to one condition
written by Christian
-
What Is The Proper Way To Say "We Calculated Fdr Values From P-Values" In A Paper?
written by Ryan Thompson
-
Identify Expressed Genes From Combined Microarray Data Sets
written by Jessica
If you don't feel comfortable answering, then suggest that the person go to a statistician. You could just comment on basics, like checking data distribution (is it binomial?) and, yes, the simple idea behind P value adjustments. Log2 fold change cut-offs is also important. If you want to go a bit more advanced, you can talk about P value distributions in the realm of quantile-quantile plots.
As bioinformaticians, many people assume that we are skilled statisticians, that we watch The Matrix all day, have the latest gadgets, et cetera.
Thank you, actually this is a question I am often being asked and each time I say only about P-value and FDR. I think they ask this question to examine how much I know in statistics.
Yes it will often be asked - no harm, though. I usually start the conversation by saying: "I'm not a statistician... but I can provide general information about which tests to employ and where".
A great combination is a bioinformatician and statistician at similar levels of their career who recognise that they can (and are eager to) learn from each other. Not every place has funding for both, though.
Thank you, even sometimes I am overwhelmed with much more tricky questions. For example if someone come to me with a list of selected genes extracted without any statistical test, what would I say about the genes :)
I would ask from where they got the list (and how)? 'Cherry picking', i.e., only seeing/choosing the things that you want to see and ignoring everything else that doesn't fit your hypothesis, still exists in research.
The best that you can do is provide adequate disclaimers/warnings about things like this (like at the top or bottom of reports and emails), and protect yourself in that way.
Let the results guide you and provide an honest feedback, complete with limitations. When evidence accumulates against a hypothesis, your hypothesis should change.