| 0 |  |
1 | |
2 | |
3 | |
4 | |
5 | |
6 | |
7 | |
8 | |
9 | |
10 | |
11 | |
12 | |
13 | |
14 | |
15 | |
16 | |
17 | |
18 | |
19 | |
A1 | |
A2 |
 |
| Module contents: |  |
| Learning objectives | |
 |
|
| What are continuous outcome data? | |
|
|
| What information do I need? | |
|
|
| Measuring the effect of treatment | |
|
|
| On skew | |
|
|
| Deciding on a change (from baseline) | |
|
|
| Next module |
What information do I need?In order to perform meta-analyses using continuous data, we require three numbers from each treatment group. These are
making six numbers in total for a two-group trial. |
|
 Read Section 8.4.3 of the Reviewers' Handbook if you need to calculate a SD from standard error, p-value or confidence intervals |
Sometimes we can readily extract these numbers from tables or the text of a report; sometimes we can't. Often it is possible to derive them from other statistics. Standard deviations are the most likely statistic to be missing in a trial report, but the Handbook includes details of how standard deviations can be obtained from standard errors, confidence intervals, t-statistics and p-values. If you have any need to perform these conversions, you should read it now. If other statistics are reported, such as medians, ranges and non-parametric tests (for example, a 'Mann-Whitney' test), then this is an indication that the outcome may have a skewed distribution. In some cases trials report nothing that will allow you to obtain a mean or a standard deviation. If this is the case you should attempt to contact the trialist and obtain the missing data. |
|
||
| © The Cochrane Collaboration 2002 | Next: Measuring the effect of treatment |