Diamond in the Forest
Asystole

Look at this diamond.
Top of the evidence pyramid.

Let's see what went into the jewellery.

Systematic review: define the question, plan the search;
Your favourite five papers need a wider body of research.
Clear eligibility: decide what can come through;
Record exclusions; check the risk of bias too.
Meta-analysis combines estimates when pooling makes sense;
Different questions in one bowl can make a very strange defence.
A forest plot gives each study an estimate and range;
Confidence intervals show uncertainty on that stage.
With inverse-variance weighting, more precision gets more say;
The specific method sets the weights along the way.
The pooled diamond has a centre and an interval to show;
You saw a shiny shape and stopped asking what we know.
A large study can be precise and still be biased;
Your spreadsheet's jeweller has a very trusting client.
Pooling flawed results can polish up the wrong direction;
A statistical blender doesn't disinfect selection.

Diamond in the forest: what's inside?
Which studies made the cut? Which results might hide?
Look at the estimates, their spread and their design;
Read the fine print underneath the shine.
Diamond in the forest: what's inside?
Which studies made the cut? Which results might hide?
Look at the estimates, their spread and their design;
Read the fine print underneath the shine.

A common-effect model posits one true effect to share;
Random effects let true effects vary there.
Estimate their mean with between-study variation;
Random effects won't fix a broken investigation.
Populations, doses, methods can make studies disagree;
Clinical differences deserve more than a shrug from me.
Heterogeneity: variation in effects across the set;
Decide what belongs together before placing any bet.
The mean can hide effects that differ across the floor;
Your 'average happy audience' has half outside the door.
Publication bias: results can shape which studies get seen;
Missing quiet findings tilt the answer on the screen.
Search registers and protocols; compare what should be there;
A missing outcome doesn't vanish just because we stare.
Funnel asymmetry has several explanations in the game;
It raises questions; it doesn't name one culprit for the blame.

Diamond in the forest: what's inside?
Which studies made the cut? Which results might hide?
Look at the estimates, their spread and their design;
Read the fine print underneath the shine.
Diamond in the forest: what's inside?
Which studies made the cut? Which results might hide?
Look at the estimates, their spread and their design;
Read the fine print underneath the shine.

I-squared: proportion of variation beyond sampling chance;
It's uncertain; study precision matters in that dance.
Tau-squared: between-study variance on the scale;
Neither tells you why the studies tell a different tale.
A prediction interval asks where another true effect might lie,
For a similar setting, with assumptions standing by.
Few studies make those estimates shaky at the knees;
Your diamond needs a setting and some honest caveats, please.

Diamond in the forest: what's inside?
Which studies made the cut? Which results might hide?
Look at the estimates, their spread and their design;
Read the fine print underneath the shine.
Diamond in the forest: what's inside?
Which studies made the cut? Which results might hide?
Look at the estimates, their spread and their design;
Read the fine print underneath the shine.

I checked every published paper.

What about the trial that never got published?

Fair. The quiet ones need a microphone.