Hello dear researchers,
1.I want to know how can I compute pooled effect size when we have three types of logarithmic Mean&SD
Type 1. Log (x) with base 10
Type 2. Ln (x)
Type 3. Just (x)
2. Should I use the formula mentioned below for bos mean and SD?!
Ln(x) = Log(x)/0.4342944819
3. Why we see positive results in meta-analysis when all included RCTs indicates null results? Is this an error?
Sincerely,
Hello Parsa,
Well, you can express the mean of log(x) or ln(x) on a scale that is similar to that of X, but neither the converted mean nor converted SD will equate to what you would have obtained using the original scores, X.
The reason is that log or ln are non-linear transforms of X. Yes, you can convert from one log scale to another (e.g., from log10 to ln, the multiplier is about 2.303), as you suggest. So, you can convert and compare from log to ln or vice versa.
However, you'll have to report results from X values separately from those using any variant of log(X) values, unless the study authors were willing to make the original data available to you.
Good luck with your work.
What are you trying to compute the effect size of? If it is just the mean of some X and some are reporting just that and some are reporting the geometric mean (base 10 for some), then as David Morse says your choices are limited since it is like the distributions vary between the loggers and non-loggers. On q3, without knowing what effect size we can't answer that.
Thanks Daniel Wright David Morse for your kind reply.
to clarify the subject I should say that the meta-analysis would be performed on on candida colonies. As you can see in the attached tables of included studies , the basis of reporting values are different (one is Ln, one is Log, and one is simple CFU/mL (that I think this one is also logarithmic). Now I want to know, how to do meta-analysis? Do you think using SMD is better? should I use both pre / post data in CMA software or post data only is OK?
Because in my analyzes, a significant effect is always shown (using SMD), while each study has shown that there is no significant difference between the two groups in all time points!
Also I have attached the result of my analysis on day 15 (with RevMan)
by the way, I am really confused due to contradictory results ( my meta-analysis and reported results in the individual studies.
Best
- Badges
- Science topic
- Similar topics
- Engineering
- Computer Engineering
- Computer