Formatting vs. Precision Problem Using Table

[Jared Greathouse]

I simulated a model 31 different times, and I want to compare how the estimator performs across various levels of Gaussian noise. - See my update at the bottom.

* Example generated by -dataex-. For more info, type help dataex
clear
input float(variance premse postmse ratio)
 .1 .31061575  .3611799 1.1627868
 .2  .3121041 .36341465 1.1644021
 .3 .29441914  .3411821 1.1588312
 .4 .29515648  .3421026 1.1590548
 .5 .29718387  .3448728 1.1604693
 .6  .2986346  .3474558 1.1634814
 .7 .30110005  .3512891 1.1666857
 .8   .303456  .3542318 1.1673254
 .9  .3064812  .3580584 1.1682882
  1  .3100766   .363158 1.1711879
1.1  .3154313  .3720172 1.1793923
1.2  .3182329  .3742517  1.176031
1.3    .32241  .3792378 1.1762595
1.4  .3291871  .3897445 1.1839602
1.5  .3338935  .3954707 1.1844217
1.6  .3461653   .412933 1.1928782
1.7  .3479584  .4143937 1.1909288
1.8  .3488469  .4131078 1.1842093
1.9  .3546743  .4203638 1.1852108
  2  .3602999   .427142 1.1855181
2.1  .3661608  .4342337 1.1859097
2.2  .3673443  .4312872 1.1740681
2.3   .377447  .4473171 1.1851122
2.4  .3957823  .4730073   1.19512
2.5  .3819886  .4414285 1.1556065
2.6  .4139824  .4952473 1.1963006
2.7  .4175476  .4990176 1.1951153
2.8  .4163164  .4958761 1.1911038
2.9  .4248772  .5064601 1.1920152
  3  .4310946  .5135548 1.1912811
6.2  .6136207  .7067071 1.1517003
end

graph hbar (asis) ratio, over(variance) ///
    bar(1, fcolor(cyan) lcolor(cyan)) ///
    blabel(bar, format(%6.3f)) ///
    ytitle(Post/Pre RMSE Ratio) //

I originally tried a bar graph, and I've also experimented with histograms, but to me, it seems like a table would be better to display this information. So, I did

* Example generated by -dataex-. For more info, type help dataex
clear
input float(variance premse postmse ratio)
 .1 .31061575  .3611799 1.1627868
 .2  .3121041 .36341465 1.1644021
 .3 .29441914  .3411821 1.1588312
 .4 .29515648  .3421026 1.1590548
 .5 .29718387  .3448728 1.1604693
 .6  .2986346  .3474558 1.1634814
 .7 .30110005  .3512891 1.1666857
 .8   .303456  .3542318 1.1673254
 .9  .3064812  .3580584 1.1682882
  1  .3100766   .363158 1.1711879
1.1  .3154313  .3720172 1.1793923
1.2  .3182329  .3742517  1.176031
1.3    .32241  .3792378 1.1762595
1.4  .3291871  .3897445 1.1839602
1.5  .3338935  .3954707 1.1844217
1.6  .3461653   .412933 1.1928782
1.7  .3479584  .4143937 1.1909288
1.8  .3488469  .4131078 1.1842093
1.9  .3546743  .4203638 1.1852108
  2  .3602999   .427142 1.1855181
2.1  .3661608  .4342337 1.1859097
2.2  .3673443  .4312872 1.1740681
2.3   .377447  .4473171 1.1851122
2.4  .3957823  .4730073   1.19512
2.5  .3819886  .4414285 1.1556065
2.6  .4139824  .4952473 1.1963006
2.7  .4175476  .4990176 1.1951153
2.8  .4163164  .4958761 1.1911038
2.9  .4248772  .5064601 1.1920152
  3  .4310946  .5135548 1.1912811
6.2  .6136207  .7067071 1.1517003
end

recast double variance

clear matrix
mkmat pre-ratio if !mi(ratio), mat(B)

format variance  %2.1g


levelsof variance, loc(vars)

cls

mat rownames B = `r(levels)'

mat l B

mata : st_matrix("r(B)", st_matrix("B"))
mata : st_matrixrowstripe("r(B)", st_matrixrowstripe("B"))


clear collect
collect get r(B)
collect layout (rowname)(colname)

Where the row names are the levels of noise I introduced. However, the rows appear to be not aligned with the format I've specified. For example, take the very last number. When I do

di %2.1g 6.199999809265137

I get

Res:
6.2

As expected. But, this isn't what the table shows, it shows the full number as though I've not reformatted it. Is this a display problem, or a precision one, and how might I do this such that the row names are rounded to a single decimal place, and the other three statistics are rounded to say, 3 decimal places?

UPDATE:
So, I've used Mata to reformat the. Now all I need to do is reformat the row names themselves

* Example generated by -dataex-. For more info, type help dataex
clear
input float(variance premse postmse ratio)
 .1 .31061575  .3611799 1.1627868
 .2  .3121041 .36341465 1.1644021
 .3 .29441914  .3411821 1.1588312
 .4 .29515648  .3421026 1.1590548
 .5 .29718387  .3448728 1.1604693
 .6  .2986346  .3474558 1.1634814
 .7 .30110005  .3512891 1.1666857
 .8   .303456  .3542318 1.1673254
 .9  .3064812  .3580584 1.1682882
  1  .3100766   .363158 1.1711879
1.1  .3154313  .3720172 1.1793923
1.2  .3182329  .3742517  1.176031
1.3    .32241  .3792378 1.1762595
1.4  .3291871  .3897445 1.1839602
1.5  .3338935  .3954707 1.1844217
1.6  .3461653   .412933 1.1928782
1.7  .3479584  .4143937 1.1909288
1.8  .3488469  .4131078 1.1842093
1.9  .3546743  .4203638 1.1852108
  2  .3602999   .427142 1.1855181
2.1  .3661608  .4342337 1.1859097
2.2  .3673443  .4312872 1.1740681
2.3   .377447  .4473171 1.1851122
2.4  .3957823  .4730073   1.19512
2.5  .3819886  .4414285 1.1556065
2.6  .4139824  .4952473 1.1963006
2.7  .4175476  .4990176 1.1951153
2.8  .4163164  .4958761 1.1911038
2.9  .4248772  .5064601 1.1920152
  3  .4310946  .5135548 1.1912811
6.2  .6136207  .7067071 1.1517003
end

clear matrix

mkmat pre-ratio if !mi(ratio), mat(B)

cls
mata
B = st_matrix("B")

B = strofreal(B, "%3.2f")

B = round(st_matrix("B"),.001)

st_matrix("B", B)

end

levelsof variance, loc(vars)


mat rownames B = `r(levels)'

mat l B

[Bjarte Aagnes]

mata : st_matrixrowstripe("r(B)", strofreal(strtoreal(st_matrixrowstripe("B")),"%2.1g"))

clear collect
collect get r(B)
collect style cell colname, nformat(%4.3f)
collect layout (rowname)(colname)

[Jared Greathouse]

Okay so here's the code as it's been updated.

* Example generated by -dataex-. For more info, type help dataex
clear *
input float(variance premse postmse ratio)
 .1 .31061575  .3611799 1.1627868
 .2  .3121041 .36341465 1.1644021
 .3 .29441914  .3411821 1.1588312
 .4 .29515648  .3421026 1.1590548
 .5 .29718387  .3448728 1.1604693
 .6  .2986346  .3474558 1.1634814
 .7 .30110005  .3512891 1.1666857
 .8   .303456  .3542318 1.1673254
 .9  .3064812  .3580584 1.1682882
  1  .3100766   .363158 1.1711879
1.1  .3154313  .3720172 1.1793923
1.2  .3182329  .3742517  1.176031
1.3    .32241  .3792378 1.1762595
1.4  .3291871  .3897445 1.1839602
1.5  .3338935  .3954707 1.1844217
1.6  .3461653   .412933 1.1928782
1.7  .3479584  .4143937 1.1909288
1.8  .3488469  .4131078 1.1842093
1.9  .3546743  .4203638 1.1852108
  2  .3602999   .427142 1.1855181
2.1  .3661608  .4342337 1.1859097
2.2  .3673443  .4312872 1.1740681
2.3   .377447  .4473171 1.1851122
2.4  .3957823  .4730073   1.19512
2.5  .3819886  .4414285 1.1556065
2.6  .4139824  .4952473 1.1963006
2.7  .4175476  .4990176 1.1951153
2.8  .4163164  .4958761 1.1911038
2.9  .4248772  .5064601 1.1920152
  3  .4310946  .5135548 1.1912811
6.2  .6136207  .7067071 1.1517003
end

mkmat pre-ratio if !mi(ratio), mat(B)

cls
mata : st_matrixrowstripe("r(B)", strofreal(strtoreal(st_matrixrowstripe("B")),"%2.1g"))

collect get r(B)
collect style cell colname, nformat(%4.3f)
collect layout (rowname)(colname)

The Mata line returns an error, however

. mata : st_matrixrowstripe("r(B)", strofreal(strtoreal(st_matrixrowstripe("B")),"%2.1g"))
    st_matrixrowstripe():  3300  argument out of range
                 <istmt>:     -  function returned error

I updated the code right as you posted, I think, so I'm not sure if I was meant to delete my Mata code or add yours as an addendum to mine. Bjarte Aagnes

[Bjarte Aagnes]

Hi, given the code example 2 in #1 the idea was to replace your

mata : st_matrixrowstripe("r(B)", st_matrixrowstripe("B"))

with

mata : st_matrixrowstripe("r(B)", strofreal(strtoreal(st_matrixrowstripe("B")),"%2.1g"))

which seems to work:

clear all

* Example generated by -dataex-. For more info, type help dataex

input float(variance premse postmse ratio)
 .1 .31061575  .3611799 1.1627868
 .2  .3121041 .36341465 1.1644021
 .3 .29441914  .3411821 1.1588312
 .4 .29515648  .3421026 1.1590548
 .5 .29718387  .3448728 1.1604693
 .6  .2986346  .3474558 1.1634814
 .7 .30110005  .3512891 1.1666857
 .8   .303456  .3542318 1.1673254
 .9  .3064812  .3580584 1.1682882
  1  .3100766   .363158 1.1711879
1.1  .3154313  .3720172 1.1793923
1.2  .3182329  .3742517  1.176031
1.3    .32241  .3792378 1.1762595
1.4  .3291871  .3897445 1.1839602
1.5  .3338935  .3954707 1.1844217
1.6  .3461653   .412933 1.1928782
1.7  .3479584  .4143937 1.1909288
1.8  .3488469  .4131078 1.1842093
1.9  .3546743  .4203638 1.1852108
  2  .3602999   .427142 1.1855181
2.1  .3661608  .4342337 1.1859097
2.2  .3673443  .4312872 1.1740681
2.3   .377447  .4473171 1.1851122
2.4  .3957823  .4730073   1.19512
2.5  .3819886  .4414285 1.1556065
2.6  .4139824  .4952473 1.1963006
2.7  .4175476  .4990176 1.1951153
2.8  .4163164  .4958761 1.1911038
2.9  .4248772  .5064601 1.1920152
  3  .4310946  .5135548 1.1912811
6.2  .6136207  .7067071 1.1517003
end

recast double variance

clear matrix
mkmat premse-ratio if !mi(ratio), mat(B)

format variance  %2.1g
levelsof variance, loc(vars)

cls
mat rownames B = `r(levels)'
mat l B

mata : st_matrix("r(B)", st_matrix("B"))
mata : st_matrixrowstripe("r(B)", strofreal(strtoreal(st_matrixrowstripe("B")),"%2.1g"))

collect get r(B)
collect style cell colname, nformat(%4.3f)
collect layout (rowname)(colname)

. collect layout (rowname)(colname)

Collection: default
      Rows: rowname
   Columns: colname
   Table 1: 31 x 3

-----------------------
    |    c1    c2    c3
----+------------------
.1  | 0.311 0.361 1.163
.2  | 0.312 0.363 1.164
.3  | 0.294 0.341 1.159
.4  | 0.295 0.342 1.159
.5  | 0.297 0.345 1.160
.6  | 0.299 0.347 1.163
.7  | 0.301 0.351 1.167
.8  | 0.303 0.354 1.167
.9  | 0.306 0.358 1.168
1   | 0.310 0.363 1.171
1.1 | 0.315 0.372 1.179
1.2 | 0.318 0.374 1.176
1.3 | 0.322 0.379 1.176
1.4 | 0.329 0.390 1.184
1.5 | 0.334 0.395 1.184
1.6 | 0.346 0.413 1.193
1.7 | 0.348 0.414 1.191
1.8 | 0.349 0.413 1.184
1.9 | 0.355 0.420 1.185
2   | 0.360 0.427 1.186
2.1 | 0.366 0.434 1.186
2.2 | 0.367 0.431 1.174
2.3 | 0.377 0.447 1.185
2.4 | 0.396 0.473 1.195
2.5 | 0.382 0.441 1.156
2.6 | 0.414 0.495 1.196
2.7 | 0.418 0.499 1.195
2.8 | 0.416 0.496 1.191
2.9 | 0.425 0.506 1.192
3   | 0.431 0.514 1.191
6.2 | 0.614 0.707 1.152
-----------------------