# Dixon test for outlier but which one is an outlier?

Following my previous question, I used Dixon test for outliers with the help of Michael Chernick answer. So now I have pvalues for say 10 numbers (basically 10 patients). But I have around 50k pvalues and I want to know that which patient has an outlier. Example-

chr     start.position    stop.pos     pvalue    pat1     pat2  pat3 ...
1              10         110            0.03      3     3      7
2              100        200            0.006   0.3     4      5


So I have around 50k rows in this format. So my first step is, I will take all the rows which have significant pvalues (lets say, less than 0.05). Ok, now I know which positions contain outliers but my next step is, which patient is an outlier?

Like in aforementioned example, I can say that for first row (first position), patient3 is an outlier and in second row, I can see patient1 is an outlier.For less number of positions, I can easily see the outlier but what if I have so many rows? Is there any way to find this thing? I am also ready for applying any other test or any other procedure to detect this.

Thanks in advance.

-
I think that if you are looking through 50000 data points you can't use these outlier detection methods. In large samples extremes are sure to occur (assuming independent cases) even when the data has a normal distribution. You have to be very careful! –  Michael Chernick May 10 '12 at 22:06
Hi, Thanks for your reply. I think I did not explain well. At a time, I am looking at 10 values and then I have p-values from dixon test from these 10 values. So I did this, for 50k rows but I applied test only on 10 values (row wise). So I know if p-value is significant (say <0.01) then I know there is a significant outlier in that row but I want to know which one (or more) is outlier. Do we have any test or procedure for this? –  Vikas May 15 '12 at 7:27