P-value to cry without tears = 0.051 | Doing a few repetitions can get a lower p-value.

P-value to cry without tears = 0.051 | Doing a few repetitions can get a lower p-value.

has been read 1469 times 2017-12-15 14:12 | Personal Category: Rna-seq| System Category: Science collection |  Key words: Rna-seq| rna-seq

This article is reproduced since the public number, has been authorized. Check out the latest articles, stay tuned for Id:rainbow-genome

Xiao Ha Source: Jian

60 Minutes Long live, 1 points wasted

P-value = 0.051 ...

Scene One: Do Rna-seq, do a few repetitions? How many times should I do? How many repetitions do you recognize when you send paper?

Scene Two: Rna-seq did 3 repetitions, with P-value < 0.05 sieve out the difference gene too little, only 2 times to sift, sieve out a lot, good Kai sen ~ ~ ~

Scenario three: The Gene KO group compared with the control group, the calculation of the phenotype after the drug treatment is significantly different. Did 3 experiments, P-value > 0.05, continue to do, do 5 repetitions, p-value < 0.05, good open sen ~ ~ ~

Reduce the number of repetitions? Increase the number of repetitions? How is that reasonable?

Take a look at the statquest video about power in statistics, which requires a few repetitions to get a lower p-value.

Unlucky Daner not only you, what happens when you increase the number of repetitions ?

After doing more repetition, 30% of the cases p-value to less than 0.05, making sense? No! It's called false positives .

It is not possible to increase repetition in order to obtain good p-value, which increases false positives.

the correct approach is to :

1. Before doing the experiment, calculate how much sample size is needed;

2. If the sample quantity is not evaluated before the experiment, the experiment is restarted after the assessment.

(You do not know the statistics of the experimental Dog's sadness)

How do you assess how many repetitions you need to make?

How many repetitions are required before an experiment,power calculation, affected by 4 factors:

The 4th one is the statistical method, the most widely used t-test has the strongest power. Focus on the first three factors:

It is best to have pre-data or published data, with a sixth sense:

If there is no upfront data and a sixth sense, there are ways:

For RNA-SEQ data:

How exactly?

To the computer, click on the bottom left corner "read the original" direct access to the G*power website. Just updated, July 2017-release 3.1.9.3, with Windows version and Mac version. Simple to use:

You may need some basic knowledge to move to a simple and understandable statquest video.

Another video of one or both tail t-test

Back to the beginning of the scene, solution:

No matter how it counts, We always use 3 patients ...

P-value to cry without tears = 0.051 | Doing a few repetitions can get a lower p-value.