Timeline for Statistical test to assess differences between two sequencing methods
Current License: CC BY-SA 3.0
20 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jul 2, 2014 at 13:07 | vote | accept | Michael | ||
| S Jul 2, 2014 at 13:07 | history | bounty ended | Michael | ||
| S Jul 2, 2014 at 13:07 | history | notice removed | Michael | ||
| Jul 1, 2014 at 1:09 | answer | added | jimmyb | timeline score: 2 | |
| Jun 30, 2014 at 11:16 | comment | added | Michael | Hey Tal, thanks for your comment! The first (test if they produce indistinguishable results) is the main goal (and the reason for my post). The second would be nice to have... | |
| Jun 30, 2014 at 9:43 | comment | added | Trisoloriansunscreen | Michael, what is the goal of this analysis? Is it to test whether the two methods produce indistinguishable results? Alternatively, is it to test whether one method is better than the other in predicting the 'ground truth' mixtures? Comparing the ratios is relevant for the former, but for the latter some error statistic has to be formed and tested. | |
| Jun 24, 2014 at 13:47 | history | tweeted | twitter.com/#!/StackStats/status/481433541361025024 | ||
| S Jun 24, 2014 at 12:27 | history | bounty started | Michael | ||
| S Jun 24, 2014 at 12:27 | history | notice added | Michael | Draw attention | |
| Jun 24, 2014 at 12:24 | history | edited | Michael | CC BY-SA 3.0 |
added 372 characters in body
|
| Jun 13, 2014 at 13:54 | comment | added | Michael | Hey Glen! Does my edit clarify things? | |
| Jun 13, 2014 at 11:17 | history | edited | Michael | CC BY-SA 3.0 |
edited body
|
| Jun 13, 2014 at 9:09 | history | edited | Michael | CC BY-SA 3.0 |
added 165 characters in body
|
| Jun 4, 2014 at 21:34 | comment | added | Glen_b | You should explain in detail the thing you are adjusting by to remove bias. The inputs to the numbers you give us. There will almost certainly be another way to deal with the problem if we know clearly what it is. At the least we may be able to derive a variance-function that works for the data you have. | |
| Jun 4, 2014 at 14:27 | comment | added | Michael | The point is that the number of seq.-reads are in the same range. So if I just normalize, is the bias I introduce acceptable? | |
| Jun 4, 2014 at 14:15 | comment | added | Glen_b | Yes, in order to work out the vriability of the proportions. It's like tossing coins. If I wonder whether two coins give the same proportion of heads, and I get the result that A gave heads 70% of the time and B gave heads 60% of the time, it makes a lot of difference whether I tossed 10 times or 1000 times. | |
| Jun 4, 2014 at 14:10 | comment | added | Michael | I have the counts! But does that make a difference? I have to normalize anyway, because the sequencing reads are proportional to the total reads per sequencing run. It's hard to tell something about absolute numbers... | |
| Jun 4, 2014 at 14:07 | comment | added | Glen_b | chi-square might work if you had counts, not just percentages | |
| Jun 4, 2014 at 13:56 | history | edited | Michael | CC BY-SA 3.0 |
edited body
|
| Jun 4, 2014 at 13:31 | history | asked | Michael | CC BY-SA 3.0 |