[OLMABA JALA]

Thanks Nick

But the code above would "only" caluculate the impact of the specific combination of charateristics that are listed in the table.What I would like to see is which of the charateristics (for instance "A") has what kind of impact on the revenue? And what impact all possible combinations have (for instance "A" with "C" or "C" with "D")?
The values A, C or D represent specific charateristics/attributes of a company. Thats why it can be the case that "C,A" appears twice (two companies with the same attributes). You are right "A,D,D" was a mistake and should be "A,C,D".

So is there a way of splitting the comma separated charateristics and compute for instance a qualitative comparative analysis in STATA? Or can I do this with another method ?

[Nick Cox]

I think you need to give a realistic data example to speed things up.

[OLMABA JALA]

Characteristic	Revenue
Headquarter in UK	287
Female CFO	284
Male CEO, Footprint in Asia, Headquarter in UK	344
Footprint in Asia	194
Footprint in Asia, Headquarter in UK, Female CFO	10
Male CEO, Female CFO	357
Headquarter in UK	212
Headquarter in UK, Female CFO	269
Footprint in Asia	248
Footprint in Asia, Headquarter in UK, Female CFO	104
Footprint in Asia	300
Male CEO, Footprint in Asia, Headquarter in UK, Female CFO	156
Footprint in Asia, Headquarter in UK, Female CFO	194
Female CFO	374
Male CEO, Female CFO	297
Headquarter in UK	5
Footprint in Asia	72
Footprint in Asia, Headquarter in UK, Female CFO	353
Headquarter in UK, Female CFO	25
Footprint in Asia	114
Male CEO, Footprint in Asia, Headquarter in UK, Female CFO	125
Headquarter in UK, Female CFO	206
Male CEO	235
Male CEO, Footprint in Asia	298
Footprint in Asia	287
Male CEO	273
Male CEO, Footprint in Asia, Headquarter in UK, Female CFO	70
Headquarter in UK, Female CFO	255
Headquarter in UK	85
Male CEO, Footprint in Asia	357
Footprint in Asia	142
Headquarter in UK	175
Footprint in Asia	145
Footprint in Asia	135
Male CEO, Footprint in Asia, Headquarter in UK	320
Footprint in Asia	311
Footprint in Asia	382
Headquarter in UK	351
Footprint in Asia, Headquarter in UK, Female CFO	88
Male CEO	244
Footprint in Asia, Headquarter in UK, Female CFO	311
Headquarter in UK, Female CFO	237
Male CEO, Footprint in Asia, Headquarter in UK, Female CFO	16
Male CEO	354
Male CEO	187
Headquarter in UK, Female CFO	314
Female CFO	376
Male CEO	89
Male CEO, Female CFO	218
Female CFO	310
Footprint in Asia	229
Footprint in Asia, Headquarter in UK, Female CFO	164
Footprint in Asia	165
Footprint in Asia	38
Headquarter in UK	305
Headquarter in UK	322
Headquarter in UK, Female CFO	368
Headquarter in UK, Female CFO	16
Male CEO, Footprint in Asia, Headquarter in UK	159
Headquarter in UK	363
Male CEO, Female CFO	117
Male CEO	43
Footprint in Asia, Headquarter in UK, Female CFO	142
Male CEO, Footprint in Asia, Headquarter in UK, Female CFO	68
Headquarter in UK	393
Footprint in Asia, Headquarter in UK, Female CFO	277
Male CEO	348
Male CEO, Female CFO	267
Headquarter in UK	344
Headquarter in UK	131
Male CEO, Female CFO	245
Male CEO	95
Male CEO, Footprint in Asia, Headquarter in UK, Female CFO	172
Male CEO, Footprint in Asia, Headquarter in UK, Female CFO	202
Male CEO, Female CFO	337
Male CEO, Footprint in Asia	147
Male CEO, Footprint in Asia	205
Male CEO, Female CFO	64
Male CEO, Footprint in Asia, Headquarter in UK, Female CFO	65
Headquarter in UK, Female CFO	288
Female CFO	257
Footprint in Asia, Headquarter in UK, Female CFO	271
Male CEO, Footprint in Asia, Headquarter in UK	108
Male CEO, Footprint in Asia, Headquarter in UK	5
Headquarter in UK	15
Headquarter in UK	204
Footprint in Asia	312
Headquarter in UK, Female CFO	223
Footprint in Asia	35
Male CEO, Footprint in Asia	225
Footprint in Asia, Headquarter in UK, Female CFO	112
Male CEO, Footprint in Asia, Headquarter in UK	195
Male CEO	2
Male CEO, Female CFO	269
Male CEO, Footprint in Asia	68
Headquarter in UK, Female CFO	110
Headquarter in UK	179
Headquarter in UK	238
Male CEO	203
Female CFO	246
Male CEO, Footprint in Asia, Headquarter in UK, Female CFO	52
Male CEO, Footprint in Asia, Headquarter in UK	45
Male CEO, Footprint in Asia, Headquarter in UK	376
Male CEO, Footprint in Asia, Headquarter in UK	59
Headquarter in UK, Female CFO	213
Male CEO, Footprint in Asia	207
Footprint in Asia, Headquarter in UK, Female CFO	88
Male CEO, Footprint in Asia, Headquarter in UK, Female CFO	127
Headquarter in UK	132
Male CEO, Female CFO	262
Male CEO, Footprint in Asia	312
Headquarter in UK	307
Footprint in Asia	67
Footprint in Asia	335
Footprint in Asia, Headquarter in UK, Female CFO	288
Male CEO, Footprint in Asia	232
Male CEO, Footprint in Asia, Headquarter in UK, Female CFO	88
Male CEO, Footprint in Asia, Headquarter in UK, Female CFO	255
Male CEO, Female CFO	250
Male CEO, Female CFO	283
Headquarter in UK	396
Male CEO	105
Male CEO, Footprint in Asia, Headquarter in UK	27
Footprint in Asia	51
Male CEO, Footprint in Asia	382
Male CEO, Footprint in Asia	101
Headquarter in UK	233
Male CEO, Female CFO	133
Footprint in Asia	331
Male CEO, Female CFO	369
Headquarter in UK, Female CFO	125
Male CEO, Female CFO	375
Male CEO, Footprint in Asia, Headquarter in UK, Female CFO	166
Male CEO, Footprint in Asia, Headquarter in UK, Female CFO	360
Female CFO	216
Headquarter in UK, Female CFO	242
Footprint in Asia, Headquarter in UK, Female CFO	244
Male CEO, Footprint in Asia	287
Male CEO, Footprint in Asia, Headquarter in UK	372
Male CEO, Footprint in Asia	218
Footprint in Asia	94
Male CEO, Footprint in Asia, Headquarter in UK, Female CFO	350
Male CEO, Footprint in Asia, Headquarter in UK	286
Male CEO, Footprint in Asia, Headquarter in UK	107
Male CEO, Footprint in Asia, Headquarter in UK	62
Footprint in Asia, Headquarter in UK, Female CFO	148
Footprint in Asia, Headquarter in UK, Female CFO	272
Male CEO, Footprint in Asia, Headquarter in UK	85
Headquarter in UK, Female CFO	276
Male CEO, Footprint in Asia, Headquarter in UK	82
Headquarter in UK, Female CFO	110
Male CEO, Female CFO	305
Male CEO, Footprint in Asia, Headquarter in UK, Female CFO	131
Footprint in Asia	369
Male CEO, Footprint in Asia, Headquarter in UK, Female CFO	345
Male CEO, Female CFO	323
Headquarter in UK, Female CFO	75
Male CEO, Female CFO	205
Footprint in Asia	1
Male CEO, Footprint in Asia, Headquarter in UK, Female CFO	317
Female CFO	372
Male CEO	353
Headquarter in UK	330
Male CEO	160
Female CFO	304
Footprint in Asia	240
Male CEO, Footprint in Asia	108
Female CFO	197
Male CEO, Footprint in Asia, Headquarter in UK	146
Male CEO, Footprint in Asia, Headquarter in UK, Female CFO	304
Male CEO	101
Female CFO	79
Footprint in Asia	119
Headquarter in UK, Female CFO	194
Female CFO	338
Footprint in Asia	314
Male CEO	137
Female CFO	91
Headquarter in UK, Female CFO	356
Footprint in Asia, Headquarter in UK, Female CFO	311
Headquarter in UK	296
Male CEO, Footprint in Asia, Headquarter in UK, Female CFO	98
Headquarter in UK, Female CFO	144
Female CFO	275
Male CEO, Footprint in Asia, Headquarter in UK	153
Male CEO, Footprint in Asia, Headquarter in UK, Female CFO	256
Footprint in Asia, Headquarter in UK, Female CFO	179
Male CEO	21
Headquarter in UK	341
Headquarter in UK	36
Male CEO, Footprint in Asia, Headquarter in UK	233
Male CEO, Footprint in Asia, Headquarter in UK	181
Headquarter in UK, Female CFO	374
Footprint in Asia	24
Male CEO, Female CFO	120
Female CFO	47
Footprint in Asia	209
Headquarter in UK, Female CFO	75
Headquarter in UK, Female CFO	188
Headquarter in UK	49
Footprint in Asia, Headquarter in UK, Female CFO	236
Male CEO	379
Male CEO, Footprint in Asia	203
Male CEO, Footprint in Asia, Headquarter in UK	120
Female CFO	306
Female CFO	202
Footprint in Asia, Headquarter in UK, Female CFO	388
Male CEO, Footprint in Asia, Headquarter in UK	254
Footprint in Asia	182
Headquarter in UK, Female CFO	239
Male CEO, Footprint in Asia	241
Footprint in Asia	101
Headquarter in UK, Female CFO	376
Footprint in Asia, Headquarter in UK, Female CFO	252
Male CEO, Female CFO	218
Footprint in Asia	13
Male CEO, Female CFO	234
Headquarter in UK, Female CFO	177
Male CEO, Female CFO	214
Headquarter in UK, Female CFO	71
Headquarter in UK	277
Headquarter in UK	102
Male CEO, Footprint in Asia	370
Male CEO, Footprint in Asia, Headquarter in UK, Female CFO	370
Female CFO	165
Male CEO, Footprint in Asia	223
Male CEO, Female CFO	236
Male CEO	33
Headquarter in UK, Female CFO	83
Male CEO	292
Footprint in Asia	46
Footprint in Asia	112
Male CEO, Footprint in Asia, Headquarter in UK, Female CFO	278
Footprint in Asia	251
Footprint in Asia, Headquarter in UK, Female CFO	287
Headquarter in UK	237
Male CEO	74
Footprint in Asia	181
Footprint in Asia, Headquarter in UK, Female CFO	18
Female CFO	143
Footprint in Asia, Headquarter in UK, Female CFO	265
Footprint in Asia	395
Male CEO, Footprint in Asia, Headquarter in UK, Female CFO	146
Footprint in Asia	275
Headquarter in UK, Female CFO	115
Headquarter in UK, Female CFO	158
Headquarter in UK, Female CFO	346
Male CEO	94
Male CEO, Footprint in Asia	222
Footprint in Asia, Headquarter in UK, Female CFO	63
Male CEO, Footprint in Asia	16
Female CFO	202
Footprint in Asia	163
Male CEO	374
Male CEO, Footprint in Asia, Headquarter in UK, Female CFO	311
Female CFO	387
Male CEO, Footprint in Asia	192
Footprint in Asia	308
Male CEO, Footprint in Asia	79
Female CFO	399
Male CEO, Footprint in Asia, Headquarter in UK	5
Male CEO	38
Male CEO, Footprint in Asia, Headquarter in UK	226
Footprint in Asia	314
Male CEO, Female CFO	144
Male CEO, Footprint in Asia, Headquarter in UK, Female CFO	120
Headquarter in UK, Female CFO	356
Footprint in Asia, Headquarter in UK, Female CFO	275
Male CEO, Footprint in Asia, Headquarter in UK	62
Headquarter in UK, Female CFO	364
Female CFO	361
Male CEO, Female CFO	163
Footprint in Asia, Headquarter in UK, Female CFO	226
Headquarter in UK	98
Headquarter in UK	349
Male CEO	136
Male CEO, Footprint in Asia, Headquarter in UK	134
Male CEO, Female CFO	130
Male CEO, Footprint in Asia	377
Male CEO, Footprint in Asia	324
Headquarter in UK	73
Footprint in Asia	272
Male CEO, Female CFO	93
Footprint in Asia	365
Headquarter in UK, Female CFO	351
Male CEO, Footprint in Asia, Headquarter in UK	228
Footprint in Asia	143
Footprint in Asia, Headquarter in UK, Female CFO	54
Female CFO	74
Footprint in Asia	33
Male CEO, Footprint in Asia, Headquarter in UK, Female CFO	153
Male CEO	118
Footprint in Asia, Headquarter in UK, Female CFO	68
Footprint in Asia	200
Male CEO, Footprint in Asia, Headquarter in UK, Female CFO	279
Male CEO, Footprint in Asia, Headquarter in UK, Female CFO	386
Male CEO	365
Male CEO, Footprint in Asia, Headquarter in UK	290
Male CEO, Footprint in Asia, Headquarter in UK, Female CFO	73

[OLMABA JALA]

Please see above the data set

[Nick Cox]

Thanks for the data example. With tabsplit from tab_chi (SSC) I see that there are just four conditions mentioned from which indicator variables can be constructed directly.

. tabsplit char, parse(,)

    Characteristic |      Freq.     Percent        Cum.
-------------------+-----------------------------------
        Female CFO |        144       24.49       24.49
 Footprint in Asia |        156       26.53       51.02
 Headquarter in UK |        152       25.85       76.87
          Male CEO |        136       23.13      100.00
-------------------+-----------------------------------
             Total |        588      100.00

. gen Asia = strpos(char, "Asia") > 0

. gen UK = strpos(char, "UK") > 0

. gen Male = strpos(char, "Male") > 0

. gen Female = strpos(char, "Female") > 0

I used groups from the Stata Journal to get an idea of joint frequency. 10 out of 16 possibilities occur in the data example.

. groups Asia UK Male Female

  +---------------------------------------------+
  | Asia   UK   Male   Female   Freq.   Percent |
  |---------------------------------------------|
  |    0    0      0        1      24      8.03 |
  |    0    0      1        0      27      9.03 |
  |    0    0      1        1      27      9.03 |
  |    0    1      0        0      31     10.37 |
  |    0    1      0        1      34     11.37 |
  |---------------------------------------------|
  |    1    0      0        0      45     15.05 |
  |    1    0      1        0      24      8.03 |
  |    1    1      0        1      29      9.70 |
  |    1    1      1        0      28      9.36 |
  |    1    1      1        1      30     10.03 |
  +---------------------------------------------+

Then you can run a model of choice. It seems that none of the indicators is strongly predictive.

. glm revenue i.Asia i.UK i.Male i.Female

Iteration 0:   log likelihood = -1829.6186  

Generalized linear models                         Number of obs   =        299
Optimization     : ML                             Residual df     =        294
                                                  Scale parameter =   12298.57
Deviance         =  3615780.674                   (1/df) Deviance =   12298.57
Pearson          =  3615780.674                   (1/df) Pearson  =   12298.57

Variance function: V(u) = 1                       [Gaussian]
Link function    : g(u) = u                       [Identity]

                                                  AIC             =    12.2717
Log likelihood   =  -1829.61865                   BIC             =    3614105

------------------------------------------------------------------------------
             |                 OIM
     revenue | Coefficient  std. err.      z    P>|z|     [95% conf. interval]
-------------+----------------------------------------------------------------
      1.Asia |  -17.61345   13.50227    -1.30   0.192    -44.07741    8.850509
        1.UK |   -16.5807   13.65573    -1.21   0.225    -43.34544    10.18405
      1.Male |  -10.93101    13.2203    -0.83   0.408    -36.84232    14.98029
    1.Female |    19.0891   13.77222     1.39   0.166    -7.903947    46.08215
       _cons |   218.9891   14.13133    15.50   0.000     191.2922     246.686
------------------------------------------------------------------------------

I fired up designplot from the Stata Journal to get an idea of descriptive statistics: See https://www.statalist.org/forums/for...riptive-tables for an overview.

 
. designplot revenue Asia UK Male Female, max(1) exclude0 variablenames ysc(r(190 220)) stat(mean median) scheme(s1color)

[OLMABA JALA]

Thank you

But by this approach one is not testing the combinations of for instance Asia and UK but only the seperate direct effects right?

[Nick Cox]

All you need to do is add interaction terms to the model.... (There are exceptions, but by and large if main effects are minor, interactions don't help, in my experience. But if interactions are what you are checking for, you just need to extend the model.)

[Oscar Ozfidan]

Olmaba
going back to your original question regarding the cluster analysis, I can say that cluster analysis is for situations (even though I am not an expert of this at all and my view is based on what I remember from school a long time ago) when you dont have an independent variable but want to classify observations into a desired number of categories. When you have a binary independent variable that is you know the class of existing observations but you want to predict the class of a new observation that you know know which class it belongs to then you need to perform discriminant analysis. If you have a dpendent variable and it is not binary like revenue, then regression is the way to go.
Looks like a part of your question involves exploring the impact of various combinations on the independent variable you have. Looking at interactions is a good and sound approach of course. In addition to that I want to point out that your case might be a good candidate for calling principal component analysis to the rescue. The principal component analysis deals with figuring out which linear combination of independent variables explains most of the variation (usually in terms of the percentage of variation explained by a specific linear combination which may not use some of the five that you are invetigating). Anyways, to me it appeared that you are investigating something that principal component analysis might be useful.

[OLMABA JALA]

Thank you!

How would you apply principal component analysis in this example in STATA?

[Nick Cox]

Sorry, but I see no point to principal component analysis (PCA) here.

What linear combination of predictors works best is precisely what regression shows you. In wanting to go beyond simple regressions, you need a principled approach to interactions and I can't see how PCA would help make that clear.

As for how would you do it: 44 posts indicates that you are not quite new to Statalist and should now be familiar with our advice in the FAQ. #3 starts

3. What should I do before I post?

Before posting, consider other ways of finding information:

• the online help for Stata
• Stata's search command, which can tell you about all built-in Stata commands, all ado-files published in the Stata Journal, all FAQs on the Stata website, www.stata.com, and user-written Stata programs available on the Internet (if you have Stata 12 or earlier, you can use findit to search all these sources at once)

and indeed

search principal component analysis

flags relevant commands. As said, I don't recommend the method myself in this case.

[Oscar Ozfidan]

Olmaba you can check the tutorial below for principal component analysis. That said my comment about various methodologies was based on the hunch that you appeared a bit lost in your original post. I would take Nick's comment that PC is not suitable in this case seriously. He is a guru on this forum.

https://jbhender.github.io/Stats506/...jects/G18.html

[OLMABA JALA]

Thank you all!!!

One last very short question:

When I calculate a regression and type in "i.groups" the results only begin with group 2
Why is group 1 not shown and how can a change that?

[Rich Goldstein]

group 1 is not shown because, if you include a constant, you must exclude one of the categories; this is sometimes called the "dummy variable trap" and you can look that up; if you want to include all categories us "ibn.group" and either "nocons" or "hascons" as an option; see

help fvvarlist
help regress

[OLMABA JALA]

One more question regarding #3:

Is this actually a usual way in research conducting a regression with a "group" variable. I am referring to "reg profit i.groups"

[Oscar Ozfidan]

Yes. when you have a group variable you must choose a base group or noconstant option to be able to get estimates. If you include the constant, the coefficient estimates will be in relation to the base group also called the "left-out" category.