Question: Plink - Quantitative Gwas With Covariates - How Do You Figure Out Intercept B0
2
jmtoung20 wrote:

I have a question regarding PLINK's linear association test, namely how do you figure out the intercept or b0 value?

I have a quantitative trait along with a number of covariates that I ran a GWAS on using the following plink command.

``````plink --bfile mydata --linear --pheno pheno.txt --covar covar.txt --pheno-name height --covar-name sex,age,race
``````

The results in 'plink.assoc.linear' looks like this for an example SNP

`````` CHR           SNP         BP   A1       TEST    NMISS       BETA         STAT            P
0   kgp22785392          0    A        ADD     1726    -0.4155       -1.537       0.1244
0   kgp22785392          0    A        sex     1726       5.59       0.7218       0.4705
0   kgp22785392          0    A        age     1726     -21.09       -2.771     0.005643
0   kgp22785392          0    A        race     1726      11.21        1.479       0.1393
``````

Plink states that the regression equation would be something like this:

``````height = b0 -0.4155*ADD + 5.59*sex -21.09*age +11.21*race + e
``````

As a sanity check, I just wanted to make sure I am understanding this equation properly.

My questions are:

1) How do I get b0? Is it just the mean height? 2) If b0 is the mean height, I assume you need to take the mean only over the values that have been included in the analysis. That is, you should only average over the same NMISS individuals that plink used. If this is the case, how do you get the exact NMISS? 3) If I plug in the values of ADD, sex, age, and race for one individual along with b0, I should be able to get the value for e. How do I check that the value I'm getting is correct?

EDIT:

I just ran a multiple regression in R and see that the values of the coefficients are very similar to the BETA values from PLINK, as expected.

R also gives the b0 value - curious why PLINK doesn't give this value?

modified 5.9 years ago by erin.dunn0220 • written 6.0 years ago by jmtoung20
0
erin.dunn0220 wrote:

I had a similar question to you - did you find a solution for PLINK to output intercepts? If you did, I would be very interested to hear. Thanks, Erin

1

I just added an option to output intercepts; using one of the builds posted at https://www.cog-genomics.org/plink2/ , replace "--linear" with "--linear intercept" and you'll get an additional "INTERCEPT" row for each variant. Let me know if you have any problems.