Group_significance and normality test

[Maria Letícia Bonatelli]

Hi,

I was looking in another questions about the comand "group_significance" and I wasn't able to find a good answer to test if my data are normally distributed or not.

I am perfoming a t_test because I am comparing two data sets, but without testing data normality, I am unable to choose between the non or the parametric_t_test. I saw something about testing the normality in the excel, but I do not know how to do that with the .biom file. Can someone help me?

Thank you in advance,

Maria

[abir...@gmail.com]

Hi, Maria,

Most OTU abundance tables are not normally distributed (see the post at https://groups.google.com/d/msg/qiime-forum/iMIAjZPJvhY/qvU11qKKrv0J ), so you may want to default to the Mann Whitney U test instead of the t-test.  QIIME does not provide any tools specifically for testing normality, so if you want to explicitly test for normality, you'll have to use another tool.  One option is to write a script in R or Python, both of which have stats libraries available.  Another, more manual option is to convert your biom to the "classic" tab-delimited format using a command like:  biom convert -i mytable.biom -o mytable.from_biom.txt --to-tsv (see http://biom-format.org/documentation/biom_conversion.html for details) and then import the result into Excel and follow advice from a source like the Cross-Validated site (for example, see https://stats.stackexchange.com/questions/72418/how-to-check-for-normal-distribution-using-excel-for-performing-a-t-test#72422 ) on how to do a normality test in Excel.

Best,

Amanda

[Maria Letícia Bonatelli]

Hi Amanda,

Thank you for the quick response! I just want to check the normality to choose the right test! You answer will help me a lot! I am gonna perform this analysis and I am sure it will work!

Thank you!

Maria

[Maria Letícia Bonatelli]

Hi Amanda,

Just a question, I was able to generate the tab-delimited format from biom. The value "0" indicates the absence of the OTU in a sample, right?

Thank you!

Maria

[abir...@gmail.com]

Yep, you're right!

Amanda

[Maria]

Hi Amanda,

I have another question regarding group_significance.py. Since core_diversity_analysis.py also returns me group_significance, I was wondering about the recomendation about filtering data before runing this script.

I acctually filtered OTU table excluding low percentage OTUs (> 0.005%) and I used rarified OTU table to perform the analysis. I though this was enough, since this script runs with core_diversity_analysis.

Is that ok?

The result seems reasonable.

Thank you again!

Maria

[Jai Ram Rideout]

Hi Maria,

core_diversity_analyses.py will rarefy your table as part of its workflow (using the rarefaction depth supplied via -e) and it will supply this rarefied table to group_significance.py. Thus, you'll want to supply your unrarefied table to core_diversity_analyses.py.

If you're running group_significance.py directly (i.e. not using core_diversity_analyses.py), you should supply your rarefied table.

Best,

Jai

[Maria]

Hi Jai,

Yeah, that is true, I used the unrarified table in the core_diversity_analyses.py. So I can use the result of group_significance.py, right?

Also, my data is not normally distributed, so I used the non parametric T_test.

Thank you for the answer!

Maria

[Jai Ram Rideout]

Hi Maria,

That sounds right. Happy QIIMEing!

Best,

Jai

[Maria]

Hi Jai!

Thanks! \o/

Best,

Maria