Forest Plots......can these be done in excel or powerpoint

  • Thread starter Thread starter Guest
  • Start date Start date
Only if you have the trees!
To be serious, please tell what is a forest plot then we might be able to
help
best wishes
 
it has a horizontal axis and the vertival axis in in the middle of that.
There is a value which is plotted as a rectangle and then a range around it
that is plotted like a SD as a line going through it...hard to describe. The
axis is also not uniform ie it goes from 0.25 to 0.50 to 1.0
 
Forest plots are an established way of presenting results of a statistical
meta-analysis. They are basicly just type of error-bar plots, with error
bars usually horizontal, (preferably) varying size of the symbol, and often
an added vertical line. To clarify with an example: typically, the symbol
would depict the odds ratio estimated from one study, the error bars would
represent the limits of its (say, 95%) confidence interval, symbol size
would reflect sample size (since a study with a larger sample ... - no space
to explain statistics here), and the line would denote the odds ratio 1
(corresponding to no effect - speaking very briefly thus simplicistically).
Different studies reviewed in the meta-analyses would be depicted one above
the other, perhaps ordered chronologically (or by the estimated OR).

Of course, the arrangement can be rotated, i.e., vertical error bars and
horizontal unity line. Also, with ORs, it is reasonable to make the numeric
axis logarithmic, while with some other effect size measure one will leave
it linear (for measures other than OR, the <no effect> line would, of
course, also placed elsewhere, usually at 0). As another point, one can use
different symbols for different directions of effect (if the topic
researched happens to be that controversial), or at least for the studies
showing stat. sig. effect vs. non-sig. ones.

Anyway, from this brief description (for more, just google on "meta-analysis
forest plot"; whoever is the original poster, he or she should first read
extensively and thoroughly on meta-analysis, anyway), it should be clear
that no special software is required for forest plots, and that Excel is
actually very useful for constructing quite useful forrest plots.

If one insists on "canned stuff", I believe the nice freeware basics stats
package Merlin
http://www.heckgrammar.kirklees.sch.uk/content/departments/science/biology/merlin.htm
does what it calls kyte-graphs, which might be used for the purpose of
producing forest plots. If not, the package and the enthousiasts who
developed it deserves credit anyway.

As for professional stats packages, NCSS (at least the latest version) does
forest plots as one of its advertised features.

Hope this helps,

Gaj Vidmar
Univ. of Ljubljana, Fac. of Medicine, Inst. of Biomedical Informatics
http://www.mf.uni-lj.si/ibmi-english [/biostat-center] [-> Software]
 
Any links to an example, so we can see one?

- Jon
-------
Jon Peltier, Microsoft Excel MVP
Peltier Technical Services
Tutorials and Custom Solutions
http://PeltierTech.com/
_______


Gaj said:
Forest plots are an established way of presenting results of a statistical
meta-analysis. They are basicly just type of error-bar plots, with error
bars usually horizontal, (preferably) varying size of the symbol, and often
an added vertical line. To clarify with an example: typically, the symbol
would depict the odds ratio estimated from one study, the error bars would
represent the limits of its (say, 95%) confidence interval, symbol size
would reflect sample size (since a study with a larger sample ... - no space
to explain statistics here), and the line would denote the odds ratio 1
(corresponding to no effect - speaking very briefly thus simplicistically).
Different studies reviewed in the meta-analyses would be depicted one above
the other, perhaps ordered chronologically (or by the estimated OR).

Of course, the arrangement can be rotated, i.e., vertical error bars and
horizontal unity line. Also, with ORs, it is reasonable to make the numeric
axis logarithmic, while with some other effect size measure one will leave
it linear (for measures other than OR, the <no effect> line would, of
course, also placed elsewhere, usually at 0). As another point, one can use
different symbols for different directions of effect (if the topic
researched happens to be that controversial), or at least for the studies
showing stat. sig. effect vs. non-sig. ones.

Anyway, from this brief description (for more, just google on "meta-analysis
forest plot"; whoever is the original poster, he or she should first read
extensively and thoroughly on meta-analysis, anyway), it should be clear
that no special software is required for forest plots, and that Excel is
actually very useful for constructing quite useful forrest plots.

If one insists on "canned stuff", I believe the nice freeware basics stats
package Merlin
http://www.heckgrammar.kirklees.sch.uk/content/departments/science/biology/merlin.htm
does what it calls kyte-graphs, which might be used for the purpose of
producing forest plots. If not, the package and the enthousiasts who
developed it deserves credit anyway.

As for professional stats packages, NCSS (at least the latest version) does
forest plots as one of its advertised features.

Hope this helps,

Gaj Vidmar
Univ. of Ljubljana, Fac. of Medicine, Inst. of Biomedical Informatics
http://www.mf.uni-lj.si/ibmi-english [/biostat-center] [-> Software]

Any links to an example, so we can see one?

- Jon
-------
Jon Peltier, Microsoft Excel MVP
Peltier Technical Services
Tutorials and Custom Solutions
http://PeltierTech.com/
_______


Claire8 wrote:


it

The
Click to expand...
Click to expand...
 
Here's an example cum description.
http://www.childrens-mercy.org/stats/weblog2005/ForestPlots.asp

Looks like a bubble chart with custom markers(square and diamond) for
the weighting and custom x error bars for the confidence intervals.

Cheers
Andy

Jon said:
Any links to an example, so we can see one?

- Jon
-------
Jon Peltier, Microsoft Excel MVP
Peltier Technical Services
Tutorials and Custom Solutions
http://PeltierTech.com/
_______


Gaj said:
Forest plots are an established way of presenting results of a
statistical
meta-analysis. They are basicly just type of error-bar plots, with error
bars usually horizontal, (preferably) varying size of the symbol, and
often
an added vertical line. To clarify with an example: typically, the symbol
would depict the odds ratio estimated from one study, the error bars
would
represent the limits of its (say, 95%) confidence interval, symbol size
would reflect sample size (since a study with a larger sample ... - no
space
to explain statistics here), and the line would denote the odds ratio 1
(corresponding to no effect - speaking very briefly thus
simplicistically).
Different studies reviewed in the meta-analyses would be depicted one
above
the other, perhaps ordered chronologically (or by the estimated OR).

Of course, the arrangement can be rotated, i.e., vertical error bars and
horizontal unity line. Also, with ORs, it is reasonable to make the
numeric
axis logarithmic, while with some other effect size measure one will
leave
it linear (for measures other than OR, the <no effect> line would, of
course, also placed elsewhere, usually at 0). As another point, one
can use
different symbols for different directions of effect (if the topic
researched happens to be that controversial), or at least for the studies
showing stat. sig. effect vs. non-sig. ones.

Anyway, from this brief description (for more, just google on
"meta-analysis
forest plot"; whoever is the original poster, he or she should first read
extensively and thoroughly on meta-analysis, anyway), it should be clear
that no special software is required for forest plots, and that Excel is
actually very useful for constructing quite useful forrest plots.

If one insists on "canned stuff", I believe the nice freeware basics
stats
package Merlin
http://www.heckgrammar.kirklees.sch.uk/content/departments/science/biology/merlin.htm

does what it calls kyte-graphs, which might be used for the purpose of
producing forest plots. If not, the package and the enthousiasts who
developed it deserves credit anyway.

As for professional stats packages, NCSS (at least the latest version)
does
forest plots as one of its advertised features.

Hope this helps,

Gaj Vidmar
Univ. of Ljubljana, Fac. of Medicine, Inst. of Biomedical Informatics
http://www.mf.uni-lj.si/ibmi-english [/biostat-center] [-> Software]

Any links to an example, so we can see one?

- Jon
-------
Jon Peltier, Microsoft Excel MVP
Peltier Technical Services
Tutorials and Custom Solutions
http://PeltierTech.com/
_______


Claire8 wrote:


it has a horizontal axis and the vertival axis in in the middle of
that.
There is a value which is plotted as a rectangle and then a range
around

it

that is plotted like a SD as a line going through it...hard to
describe.

The

axis is also not uniform ie it goes from 0.25 to 0.50 to 1.0
Click to expand...
Click to expand...
Click to expand...
 
Oh, that's not too bad.

Bubble chart with custom markers: make the bubble chart, draw the
desired shape in the worksheet, copy it, select the series, and paste.
Wow, square bubbles. The diamond is also a custom marker. See also

http://peltiertech.com/Excel/ChartsHowTo/CustomMarkers.html

Custom error bars:

http://peltiertech.com/Excel/ChartsHowTo/ErrorBars.html

The dashed vertical line is an error bar on the diamond. the vertical
line at 1 can either be the category axis crossing at 1, or an error bar
on an invisible point. A bubble chart can't have an XY series as part
of a combination chart, as I like to do in my charts, but you can use a
bubble chart series for this with no border and no fill. Otherwise, any
of these error bar techniques for adding a line would work fine:

http://peltiertech.com/Excel/Charts/AddLine.html

- Jon
-------
Jon Peltier, Microsoft Excel MVP
Peltier Technical Services
Tutorials and Custom Solutions
http://PeltierTech.com/
_______
 
Back
Top