How do I create a new table variable based on the concatenation of two or more categorical variables?

I am looking to simply concatenate three variables in a table (2310 rows), where those variables are categorical. I tried:

my_table.new_variable = strcat(my_table.variable1, "_", my_table.variable2, "_", my_table.variable3);

but it does not like the delimiter or the categories, and gave an error.

Error using strcat
Inputs must be character vectors, cell arrays of character vectors, or string arrays.

I tried

a = [my_table.variable1, my_table.variable2, my_table.variable3];
my_table.new_variable = strjoin(a , "_");

but these are categorical and maybe that is why this error (maybe same as above)? a is 2310x3 categorical

Error using strjoin
First input must be a string array or cell array of character vectors.

What I have are three categorical variables that I use for grouping, and I want to create charts and summary statistics by various combinations of the grouping variables. An example would be

my_table.make - [ Ford, Toyota, BMW]

my_table.types - [SUV, Sedan, Compact ]

I want

my_table.configuration - [ Ford_SUV, Ford_Sedan, BMW_Compact, ... etc ].

However, my data set does not have all combinations, which is why I first thought of strcat(). Maybe I am missing a feature in the use of groups in charts and stats.

Answer 1

[Steven Lord]

Do you want these new identifiers (Ford_SUV, Ford_Sedan, BMW_Compact, etc.) to be new categories in a new categorical variable in your table or do you want them to be text labels ("Ford_SUV", "Ford_Sedan", "BMW_Compact", etc.)?

If the latter:

c = categorical(["Red", "Blue", "Green"])

c = 1×3 categorical array
     Red      Blue      Green 

s = string(c(1)) + "_" + string(c(3))

s = "Red_Green"

Or for a vector case:

whichColors = randi(numel(c), 5, 3);
sv = string(c(whichColors))

sv = 5×3 string array
    "Green"    "Red"     "Blue" 
    "Blue"     "Red"     "Blue" 
    "Blue"     "Red"     "Blue" 
    "Green"    "Blue"    "Green"
    "Green"    "Red"     "Blue" 

y = join(sv(:, [1 3]), "_")

y = 5×1 string array
    "Green_Blue"
    "Blue_Blue"
    "Blue_Blue"
    "Green_Green"
    "Green_Blue"

If the latter it gets a little trickier to work with, depending on what you're trying to do. You won't be able to easily refer to the new categorical array using the categories from the old one even if one of the new categories was created from an instance of that category in the original categorical array.

c2 = categorical(y)

c2 = 5×1 categorical array
     Green_Blue 
     Blue_Blue 
     Blue_Blue 
     Green_Green 
     Green_Blue 

joinedCategories = categories(c2)

joinedCategories = 3×1 cell array
    {'Blue_Blue'  }
    {'Green_Blue' }
    {'Green_Green'}

originalCategories = categories(c)

originalCategories = 3×1 cell array
    {'Blue' }
    {'Green'}
    {'Red'  }

y == originalCategories(1)

ans = 5×1 logical array
   0
   0
   0
   0
   0

[Ted H]

I am trying to achieve the former. "new identifiers (Ford_SUV, Ford_Sedan, BMW_Compact, etc.) to be new categories in a new categorical variable in your table"

the join recommendation yeilds this error message

Error using join
First argument must be text.

I want to run summary statistics for a Ford SUV and compare to a BMW Compact, or create a box whiskers plot with both of these (and the other 9 configurations) on the same chart. I can easily get this information for Ford, or for SUV, etc. I would like to combine.

I suppose one way would be to create a non-catetorical variable based on the categorical, then use strjoin(). But I can't be the first person to what to look at combinations of categories (not just filtering/indexing) by multiple categories). So I was thinking there is any easy way to do this. strjoin() should work, but many matlab tools become unavailable when a table variable becomes categorical.