Timeline for How can I combine data from 2 separate experiments?
Current License: CC BY-SA 3.0
8 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Oct 2, 2013 at 16:20 | comment | added | Brenda | Thanks everyone! To clarify: I performed experiment 1 (E1) using 10 animals (5/group) and measured protein levels. I performed a t-test (using graphpad prism) which told me that the variance between groups was sign. different. I then re-analyzed, t-test + welch. I then repeated with 10 new animals. Neither experiment alone was sign. The trends are the same (treatment reduced protein). However, the levels in both groups (treated and untreated ) are higher in E1 than in E2. If I combine experiments, my error bars are huge. So, my question is, can I combine the data from both experiments? | |
| Oct 1, 2013 at 23:46 | comment | added | John | I think the critical thing is not that these are separate experiments but that t-tests have already been done. | |
| Oct 1, 2013 at 23:36 | answer | added | John | timeline score: 3 | |
| Oct 1, 2013 at 22:51 | comment | added | Glen_b | "I performed a t-test with welch correction (because there was significant variance)" -- this is not an advisable approach. Unless you have - a priori - reason to think that the variances will be similar, it's better (in terms of your tests having close to their nominal properties) to just use the Welch procedure, rather than use a hypothesis test to decide when to use it. | |
| Oct 1, 2013 at 22:03 | review | First posts | |||
| Oct 1, 2013 at 22:05 | |||||
| Oct 1, 2013 at 21:59 | comment | added | AdamO | In what sense were these experiments separate then? | |
| Oct 1, 2013 at 21:54 | comment | added | whuber♦ | Would it be correct to assume your combined data could be accurately and completely represented by a collection of ordered triples of the form $(i, x, y)$ where $i$ identifies the experiment, $x$ indicates whether treatment was applied or not, and $y$ is the protein level? Note that this implies there exist no other physical relationships among the data: no two triples refer to the same animal, animals were not housed together, they aren't groups of siblings, and so on. | |
| Oct 1, 2013 at 21:45 | history | asked | Brenda | CC BY-SA 3.0 |