x-axis scale for metan

[Jalal Choupani]

Hi all,
I am trying to create a forest plot using metan with a standardized scale of -24, 0, 1, 2, 5, 10, 41, 65. I want to set equidistant space between these ticks. Stata uses the given points on the x-axis with a linear scale, but for example, I want to have equal space between (-24-0) and (0-1) and etc.
I appreciate any suggestion to deal with this issue.

Here is the code

metan RR LRR URR, label(namevar=study) random effect(RR) xlab(-24, 0, 1, 2, 5, 10, 41, 65)

[Leonardo Guizzetti]

Are you peroeming a meta-analysis or trying to use the forest plotting function of metan? If this is actually for a meta-analysis, the choice of scale is bizarre to me and does not have an immediately obvious advantage. I would suggest not changing the scale here.

If instead you want to mimic the function of the forest plot, you can achieve this with a combination of -twoway rapike- and overlaid with -twoway scatter-. The first lets you plot the whiskers/intervals and the second lets you plot the point estimate. The one thing you will need to do is to manipulate you dataset to have discrete levels for each of your xaxis points (e.g., 1, 2, 3,...) and the label those points with the desired labels using the appropriate xscale() option in your graphing command.

I'm on mobile right now so I cannot offer a working example, but this should give you something to get started.

[Jalal Choupani]

Dear Leonardo,
Yes, I am performing a meta-analysis. The problem is that the difference between the confidence intervals for the various studies is very large and in the final shape of the forest plot, small CI in the majority of studies is not well presented.
as follows:

here is my data:

 dataex treatment logrr selogrr

----------------------- copy starting from the next line -----------------------


	
	
* Example generated by -dataex-. To install: ssc install dataex
clear
input str26 treatment float(logrr selogrr)
"1"  2.15 1.03
"2"  -.35  .12
"3"  -.43  .17
"3"  -.05  .22
"3" -1.36  .37
"3"  -.71  .34
"3"  -.38  .19
"4"  1.18  .33
"5"   .24  .23
end
------------------ copy up to and including the previous line ------------------

Listed 9 out of 9 observations

. gen RR=exp(logrr)

.
. gen LRR=exp(logrr-1.96*selogrr)

.
. gen URR=exp(logrr+1.96*selogrr)
metan RR LRR URR, by(treatment) random effect(RR) xlab(.1, 1, 5, 64) xline(1) xsize(10) ysize(8)

           Study     |     ES    [95% Conf. Interval]     % Weight
---------------------+---------------------------------------------------
     1
1                    |    8.585     1.140    64.638          0.00
 Sub-total           |
  D+L pooled ES      |    8.585   -23.164    40.334          0.00
---------------------+---------------------------------------------------
     2
2                    |    0.705     0.557     0.892         18.14
 Sub-total           |
  D+L pooled ES      |    0.705     0.537     0.872         18.14
---------------------+---------------------------------------------------
     3
3                    |    0.651     0.466     0.908         16.68
4                    |    0.951     0.618     1.464         11.09
5                    |    0.257     0.124     0.530         17.18
6                    |    0.492     0.252     0.957         12.90
7                    |    0.684     0.471     0.992         15.52
 Sub-total           |
  D+L pooled ES      |    0.579     0.355     0.804         73.37
---------------------+---------------------------------------------------
     4
8                    |    3.254     1.704     6.214          0.81
 Sub-total           |
  D+L pooled ES      |    3.254     1.000     5.509          0.81
---------------------+---------------------------------------------------
     5
9                    |    1.271     0.810     1.995          7.68
 Sub-total           |
  D+L pooled ES      |    1.271     0.679     1.864          7.68
---------------------+---------------------------------------------------
Overall              |          
  D+L pooled ES      |    0.680     0.472     0.887        100.00
---------------------+---------------------------------------------------
 Heterogeneity calculated by formula
  Q = SIGMA_i{ (1/variance_i)*(effect_i - effect_pooled)^2 }
where variance_i = ((upper limit - lower limit)/(2*z))^2

Test(s) of heterogeneity:
               Heterogeneity  degrees of
                 statistic     freedom      P    I-squared**   Tau-squared

1                    0.00          0         .         .%       0.0000
2                    0.00          0         .         .%       0.0000
3                   13.31          4      0.010     69.9%       0.0440
4                    0.00          0         .         .%       0.0000
5                    0.00          0         .         .%       0.0000
Overall             26.56          8      0.001     69.9%       0.0545
** I-squared: the variation in ES attributable to heterogeneity)

Note: between group heterogeneity not calculated;
only valid with inverse variance method

Significance test(s) of ES=0

1                     z=  0.53     p = 0.596
2                     z=  8.26     p = 0.000
3                     z=  5.06     p = 0.000
4                     z=  2.83     p = 0.005
5                     z=  4.20     p = 0.000
Overall               z=  6.42     p = 0.000
-------------------------------------------------------------------------