Exercise: Prophylactic administration of parenteral steroids, answer 7

Question 7: In Figure 2, why are there arrowheads on the lines for Francois and for Gaussorgues?

Suggested answer:

These lines represent the 95% confidence intervals for the estimated odds ratios for these two trials. Arrowheads are included when the scale on the horizontal axis is not wide enough to show the full confidence interval.

For Francois, the confidence interval is 0.01 to 0.51. The horizontal scale goes down to 0.01 and so should be able to show it. The line need go only to the edge of the graph, at 0.01. In fact, if we increase the number of decimal places, we find that the large sample approximation 95% confidence interval for the odds ratio is 0.0087 to 0.5077. These round to 0.01 and 0.51, as in the paper, but I suspect that the graph-drawing progam works to more decimal places that it presents and so put the arrowhead on.

For Gaussorgues, the confidence interval is 0.24 to 106.65. The horizontal scale would have to go up to above 100 to show it, i.e. the scale length would be increased.

Rather than have a graph which would be mostly empty space, below 0.01 and above 100, the arrows have been added to indicate that the confidence interval extends beyond the frame.

At least the point estimates are included.

Back to question.

Back to Systematic Reviews index.

To Martin Bland's home page.

This page maintained by Martin Bland.
Last updated: 18 February, 2009.

Back to top.