Question 11: How could we check for any evidence of publication bias?
To do this we must click "Return to table" again. There is an option "Analyses" at the top of the screen. This contains the option "Publication bias"
This looks quite symmetrical, so there is nothing here to suggest publication bias.
Clicking "Table" gives a large number of analyses which have been proposed to test for publication bias.
I clicked "next table" until I came to the Begg and Mazumdar rank correlation, as covered in the lectures. I chose the bottom P value, which is 0.46551, i.e. P = 0.5, no evidence of any publication bias.
I continued through the screen and found this:
| Duval and Tweedie's trim and fill | ||||||||
|---|---|---|---|---|---|---|---|---|
| Fixed Effects | Random Effects | Q Value | ||||||
| Studies Trimmed | Point Estimate | Lower Limit | Upper Limit | Point Estimate | Lower Limit | Upper Limit | ||
| Observed values | 0.67581 | 0.53524 | 0.85330 | 0.66815 | 0.47022 | 0.94939 | 12.76269 | |
| Adjusted values | 0 | 0.67581 | 0.53524 | 0.85330 | 0.66815 | 0.47022 | 0.94939 | 12.76269 |
Trim and fill could find no studies to put in or take out, providing more evidence that there is no evidence of publication bias. (There may still be publication bias, but we have nothing to suggest it here.)
Back to question.
Back to Systematic Reviews index.
This page maintained by Martin Bland.
Last updated: 19 February, 2010.
