Timeline for MicroArray, tesing if a sample is the same with high variance data.
Current License: CC BY-SA 2.5
19 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Feb 19, 2010 at 17:35 | answer | added | Andrew Mullhaupt | timeline score: 2 | |
| Feb 17, 2010 at 21:37 | vote | accept | Lisa | ||
| Feb 17, 2010 at 15:25 | answer | added | fedja | timeline score: 2 | |
| Feb 17, 2010 at 5:44 | history | edited | Lisa | CC BY-SA 2.5 |
added 155 characters in body
|
| Feb 17, 2010 at 5:43 | comment | added | Lisa | @sheldon-cooper answer, "Note: neither of this will give a "significance score"" A confidence value is good enough. I don't need a statistical "Test" | |
| Feb 17, 2010 at 5:24 | answer | added | user3035 | timeline score: 1 | |
| Feb 17, 2010 at 5:23 | comment | added | Lisa | DNA chemical binding at a probe is not discrete. There is a strength to the binding. In addition some probes "overlap" in that a specific DNA sequence will bind perfectly to one but well to the adjacent(not physically, genetically adjacent) probes. | |
| Feb 17, 2010 at 5:13 | comment | added | Lisa | This sounds right. The sample/DNA that the array is matched to "should" give the highest readings at each probe as it will chemically bind perfectly. There error is in the measurement of this. So it is common for a non_matching sample to have a higher measurement at a probe even though it did not bind perfectly. Again this is due to measurement error. While we have a "training set" the goal is this would be used to find the best set of helper Xi for each probe. | |
| Feb 17, 2010 at 5:04 | comment | added | fedja |
All right. So, if we subtract $M_i$ and divide by $\sigma_i$, the question boils down to whether a given set of measurements $Y_i$ can come from the independent samples of the standard normal law. The difficulty is that we want to detect the (possibly present) non-zero means that are of size less than $1$, so each individual measurement doesn't tell us much. Also, we do not know if the means are positive or negative (Or do we? Is the mean measurement for a non-matching sample always below the one for a matching one? If so, this may help a lot)
|
|
| Feb 17, 2010 at 4:52 | comment | added | Lisa | @fedja, "given set Xi of numeric measurements" Actually for a specific Xa is from (Ma, Sa) but there is a set of xi that can help. I need a method to find the helper Xi set and then a method for them to help test Xa. Does that make sense. Maybe this: given collection of means Mi and standard deviations Si (Mi,Si). Want to find a set Xi that can help check the hypothesis Xa of numeric measurement can be from (Ma,Sa). (not sure how to enter the notation you do) | |
| Feb 17, 2010 at 4:39 | history | edited | Lisa | CC BY-SA 2.5 |
added 88 characters in body
|
| Feb 17, 2010 at 4:32 | comment | added | fedja |
I see. It seems like you are saying that we have a given collection of means $m_i$ and standard deviations $\sigma_i$ and want to check the hypothesis that a given set $X_i$ of numeric measurements can be a sample of independent normal random variables with those means and standard deviations. You want to make sure that if it is, then the probability that you would say that it isn't is below 5% but you would still like to say that it isn't if it comes from the normal distributions where plenty of means differ from the conjectured ones noticeably. Is this translation correct?
|
|
| Feb 17, 2010 at 4:31 | history | edited | Lisa | CC BY-SA 2.5 |
Added an different example; added 2 characters in body
|
| Feb 17, 2010 at 4:12 | comment | added | Lisa | @Fedja, yes, Let me try an example, lets say I have 10 measurements (1-10) for the for the base DNA I have stats for each measurement location. Now I take an unknown sample and find the measurements. The variance in the base data is such that almost any sample at the individual measurement level will be in the 95% confidante int at the measurement point. I know that some of the other measurement points can help but I need a method to identify the right ones and a method to use them once identified. Maybe I will try to edit the original post with a no bio example. | |
| Feb 17, 2010 at 4:04 | comment | added | Lisa | I have several samples that have been exposed to the dna that matches the array. But again the variance is large. In addition you can't say that with the right dna the probe should have a reading of say 100 or some specific measurement. | |
| Feb 17, 2010 at 4:02 | comment | added | fedja | "There is a very high variance at each probe across the same dna. So it is very difficult to know if a individual sample matches (same dna) at the probe level". What exactly do these sentences mean? Do you mean that even if there is a match, you may easily fail to detect it and come up with a wrong result on each particular probe or you mean something else? | |
| Feb 17, 2010 at 3:52 | comment | added | Douglas Zare | Do you have a model for how the results should look if you know how the sample DNA compares with the array's DNA? | |
| Feb 17, 2010 at 3:42 | history | edited | Lisa | CC BY-SA 2.5 |
added 173 characters in body
|
| Feb 17, 2010 at 3:29 | history | asked | Lisa | CC BY-SA 2.5 |