Wednesday, July 8, 2009

Data outliers....

We had an interesting discussion this morning in a group meeting. Say you do an experiment on 5 independent days. Each day's data consists of multiple parallel replicates for each experiment, including an internal positive control. If one of the days shows that the internal control was actually negative, is it ok to throw away the whole data set for that day (because something was wrong with the assay?)? What if the other days showed a lot (or very little) variation? Does it make any difference if the experiment is a Western Blot, versus an immunostaining, versus a quantitative measurement (like cell proliferation values, or a luciferase reporter)? To be honest, I think I'd throw out that whole day as an off day and repeat it or use the n of 4 days... What say you?

In other news, I got wicked nice looking data with my new antibodies yesterday... today I'll try to reconfirm with antibodies against the tags instead of the protein of interest. AND I think there might be someone to take over my project when I leave, and s/he'd be the PERFECT choice - I'd be so happy to have this settled and start some work together before I have to leave. So, yeah!

No comments:

Post a Comment