Weird issue with constraint in 'taxes' data

[leeper]

Please specify whether your issue is about:

• a possible bug
• a question about package functionality
• a suggested code or documentation change, improvement to the code, or feature request

Something weird going on here:

> amce(taxes, chose_plan ~ taxrate2 * taxrev)
Waiting for profiling to be done...
Error in coefs_tmp[!aliased, ] <- coef_summary : 
  number of items to replace is not a multiple of replacement length
> traceback()
2: get_coef_summary(mod = mod, data = data, id = id, alpha = alpha) at amce.R#135
1: amce(taxes, chose_plan ~ taxrate2 * taxrev)
> cj(taxes, chose_plan ~ taxrate2 * taxrev, estimate = "amce")
Waiting for profiling to be done...
Error in coefs_tmp[!aliased, ] <- coef_summary : 
  number of items to replace is not a multiple of replacement length
> traceback()
3: get_coef_summary(mod = mod, data = data, id = id, alpha = alpha) at amce.R#135
2: amce(data = data, formula = formula, id = id, weights = weights, 
       feature_order = feature_order, feature_labels = feature_labels, 
       level_order = level_order, ...) at cj.R#89
1: cj(taxes, chose_plan ~ taxrate2 * taxrev, estimate = "amce")

Seems to only occur with this feature combination, but not without constraint:

> amce(taxes, chose_plan ~ taxrate2 + taxrev)
Waiting for profiling to be done...
     outcome statistic                      feature       level    estimate   std.error         z            p        lower       upper
1 chose_plan      amce Tax rate for $10,000-$35,000  10-35k: 5%  0.00000000          NA        NA           NA           NA          NA
2 chose_plan      amce Tax rate for $10,000-$35,000 10-35k: 15%  0.02907463 0.008612538  3.375849 7.358823e-04  0.012194365  0.04595489
3 chose_plan      amce Tax rate for $10,000-$35,000 10-35k: 25%  0.01679525 0.010284125  1.633124 1.024430e-01 -0.003361265  0.03695176
4 chose_plan      amce Tax rate for $10,000-$35,000 10-35k: 35% -0.04091056 0.011919199 -3.432324 5.984314e-04 -0.064271757 -0.01754936
5 chose_plan      amce                  Tax revenue        <75%  0.00000000          NA        NA           NA           NA          NA
6 chose_plan      amce                  Tax revenue      75-95% -0.04588052 0.008989190 -5.103965 3.326091e-07 -0.063499005 -0.02826203
7 chose_plan      amce                  Tax revenue     95-105% -0.05998096 0.011307260 -5.304641 1.128948e-07 -0.082142777 -0.03781913
8 chose_plan      amce                  Tax revenue    105-125% -0.07391775 0.011029219 -6.701994 2.055952e-11 -0.095534623 -0.05230088
9 chose_plan      amce                  Tax revenue       >125% -0.12287030 0.012610874 -9.743202 1.972386e-22 -0.147587158 -0.09815344

Doesn't happen with unconstrained features:

> cj(taxes, chose_plan ~ taxrate1 * taxrev, estimate = "amce")
Waiting for profiling to be done...
     outcome statistic               feature     level     estimate   std.error           z            p        lower        upper
1 chose_plan      amce Tax rate for <$10,000  <10k: 0%  0.000000000          NA          NA           NA           NA           NA
2 chose_plan      amce Tax rate for <$10,000  <10k: 5%  0.006844784 0.008657404   0.7906278 4.291612e-01 -0.007395378  0.021084946
3 chose_plan      amce Tax rate for <$10,000 <10k: 15% -0.050383031 0.008578737  -5.8730128 4.279453e-09 -0.064493798 -0.036272265
4 chose_plan      amce Tax rate for <$10,000 <10k: 25% -0.167008614 0.009084111 -18.3846949 1.742172e-75 -0.181950648 -0.152066581
5 chose_plan      amce           Tax revenue      <75%  0.000000000          NA          NA           NA           NA           NA
6 chose_plan      amce           Tax revenue    75-95% -0.002781705 0.009200863  -0.3023309 7.623998e-01 -0.017915778  0.012352367
7 chose_plan      amce           Tax revenue   95-105% -0.011051298 0.010595316  -1.0430362 2.969315e-01 -0.028479042  0.006376446
8 chose_plan      amce           Tax revenue  105-125% -0.034789897 0.009172065  -3.7930278 1.488215e-04 -0.049876602 -0.019703193
9 chose_plan      amce           Tax revenue     >125% -0.072428072 0.010124390  -7.1538211 8.439503e-13 -0.089081211 -0.055774933

Doesn't occur with immigration:

> cj(immigration, ChosenImmigrant ~ Education * Job, id = ~ CaseID)
           outcome statistic                feature               level     estimate  std.error          z            p        lower      upper
1  ChosenImmigrant      amce Educational Attainment           No Formal  0.000000000         NA         NA           NA           NA         NA
2  ChosenImmigrant      amce Educational Attainment           4th Grade  0.032242795 0.01564257  2.0612208 3.928198e-02  0.006513054 0.05797254
3  ChosenImmigrant      amce Educational Attainment           8th Grade  0.055971117 0.01568247  3.5690243 3.583132e-04  0.030175749 0.08176649
4  ChosenImmigrant      amce Educational Attainment         High School  0.117940436 0.01569985  7.5122007 5.814154e-14  0.092116478 0.14376439
5  ChosenImmigrant      amce Educational Attainment    Two-Year College  0.164346505 0.02408703  6.8230296 8.914028e-12  0.124726870 0.20396614
6  ChosenImmigrant      amce Educational Attainment      College Degree  0.191558995 0.02414755  7.9328533 2.141674e-15  0.151839805 0.23127818
7  ChosenImmigrant      amce Educational Attainment     Graduate Degree  0.178682719 0.01774300 10.0706030 7.451979e-24  0.149498080 0.20786736
8  ChosenImmigrant      amce                    Job             Janitor  0.000000000         NA         NA           NA           NA         NA
9  ChosenImmigrant      amce                    Job              Waiter -0.004555776 0.01742414 -0.2614635 7.937351e-01 -0.033215933 0.02410438
10 ChosenImmigrant      amce                    Job Child Care Provider  0.013684131 0.01745927  0.7837744 4.331725e-01 -0.015033819 0.04240208
11 ChosenImmigrant      amce                    Job            Gardener  0.015863618 0.01760330  0.9011731 3.674963e-01 -0.013091228 0.04481846
12 ChosenImmigrant      amce                    Job   Financial Analyst  0.065638780 0.03093098  2.1221049 3.382893e-02  0.014761848 0.11651571
13 ChosenImmigrant      amce                    Job Construction Worker  0.033689581 0.01758459  1.9158585 5.538311e-02  0.004765508 0.06261365
14 ChosenImmigrant      amce                    Job             Teacher  0.074107007 0.01752927  4.2276146 2.361819e-05  0.045273920 0.10294009
15 ChosenImmigrant      amce                    Job Computer Programmer  0.078108991 0.03011503  2.5936881 9.495257e-03  0.028574178 0.12764380
16 ChosenImmigrant      amce                    Job               Nurse  0.089129556 0.01748420  5.0977211 3.437667e-07  0.060370613 0.11788850
17 ChosenImmigrant      amce                    Job  Research Scientist  0.135076380 0.03014113  4.4814640 7.413274e-06  0.085498637 0.18465412
18 ChosenImmigrant      amce                    Job              Doctor  0.166308291 0.03004119  5.5360091 3.094415e-08  0.116894933 0.21572165

What's going on here?

[leeper]

Unstated constraint in the design?

> subset(cj_props(taxes, ~ taxrate2 + taxrev), Proportion == 0)
      taxrate2 taxrev Proportion
4  10-35k: 35%   <75%          0
17  10-35k: 5%  >125%          0

Maybe so but why doesn't the interaction capture this? Probably because taxrev == "<75%" is the reference category.