What Salmoneus said about possessive classifiers is relevant, and there's a chapter and a
feature in WALS.info about it.
Also, if a language has inalienable (or direct) possession and also has obligatorily possessed (or dependent) nouns, inalienably possessed items
tend to be obligatorily possessed, and obligatorily possessed items
tend to be inalienably possessed.
About where it is marked; see
Chapter 24 of WALS.info. The marking may all be on the classifier, or on some other word that signifies there is a possessor/possession relationship, marked with syntactic and/or semantic information about the possessor and the possession. That's covered in the "1.5 Other".
There's more in
Chapter 57 of WALS.info; particularly about affixes that tell about a pronominal possessor.
Chapter 58 of WALS.info talks about obligatory possessive inflection. It includes stuff about "bound inalienables", and about non-possessible nouns, and about obligatory use of possessive classifiers to show semantic possession of a grammatically-nonpossessible noun.
(Chapter 117 also
may have stuff you might find relevant to your questions.)
WALS.info lets you combine up to four features together.
Let's combine features 24A, 58A, 58B, and 59A.
Edit: Chichimeca-Jonaz has head-marking in possessive noun phrases, has obligatory possessive inflection (so, dependent nouns), has two to four possessive nouns (so, non-possessible nouns), and has more than five classes of possessive classification.
It is the only language sampled in WALS.info's database recorded with a value other than "None reported" for "Number of Possessive Nouns" (Feature 58B), and a value other than "No marking" for "Locus of Marking in Possessive Noun-Phrases" (Feature 24A), and a value of "Exists" for "Obligatory Possessive Inflection" (Feature 58A).
Edit: The following languages have interesting combinations of values of features 24A, 58A, and 59A:
Code: Select all
Two classes / Double marking / Exists
Dani (Lower Grand Valley)
Three to five classes / Head marking / Exists
Wichí
Three to five classes / Double marking / Exists
Burushaski
More than five classes / Head marking / Exists
Chichimeca-Jonaz
Two classes / Dependent marking / Exists
Wintu
Khoekhoe
Two classes / Head marking / Exists
Kui (in Indonesia)
Tanglapui
Hatam
Quileute
Tunica
Haida
Barasano
Koasati
Edit: The following languages have interesting combinations of values for features 24A, 58B, and 59A:
Code: Select all
Two classes / One / Head marking
Washo
Three to five classes / One / Dependent marking
Trumai
Three to five classes / Five or more / Double marking
Kipea
More than five classes / Two to four / Head marking
Chichimeca-Jonaz
Edit: The following languages have interesting combinations of values for features 58A, 58B, and 59A:
Code: Select all
Three to five classes / Two to four / Exists
Paamese
More than five classes / Two to four / Exists
Chichimeca-Jonaz
Edit: So that's eighteen natlangs that might be interesting from the point-of-view of your question.
Barasano
Burushaski
Chichimeca-Jonaz
Dani (Lower Grand Valley)
Haida
Hatam
Khoekhoe
Kipea
Koasati
Kui (in Indonesia)
Paamese
Quileute
Tanglapui
Trumai
Tunica
Washo
Wichí
Wintu
It might be more informative, at least at first, to combine them in pairs:
http://wals.info/combinations/58A_59A#2/25.5/153.7 has seven combinations with more than one language each, and one combination with 1 language.
http://wals.info/combinations/24A_58A#2/25.5/157.5 has nine combinations with more than one language each, and one combination with 0 languages.
http://wals.info/combinations/58A_58B#2/25.5/153.7 has six combinations with more than one language each, and two combinations with 0 languages.
http://wals.info/combinations/24A_59A#2/25.5/157.5 has 15 combinations with more than one language each, two combinations with 1 language each, and three combinations with 0 languages.
http://wals.info/combinations/58B_59A#2/25.5/153.7 has five combinations with more than one language each, seven combinations with 1 language each, and four combinations with 0 languages.
http://wals.info/combinations/24A_58B#2/25.5/157.5 has seven combinations with more than one language each, two combinations with 1 language each, and 11 combinations with 0 languages.
For most of those pairs of features, there are more combinations of values for which there is at least one language sampled in their database, than one would expect (IMO).
The exception seems to be 24A_58B.
If you combine them three at a time you see an awful lot of zeroes, which means no language in their sample database was recorded with that particular combination of values for those three features.
http://wals.info/combinations/24A_58A_59A#2/25.5/157.5 has at least one language in most combinations. 12 combinations have 0 languages; the other 28 combinations have at least one each. Just over two-thirds the possible combinations actually occur in at least one language in WALS.info's database.
http://wals.info/combinations/59A_58B_58A#2/25.5/153.7 has 17 zeroes and 15 non-zeroes. Just under half the possible combinations actually occur in at least one language in WALS.info's database.
The remaining threesomes contain both 24A and 58B, and so more than half their combinations of values have 0 languages in WALS.info's database recorded as reported with those combination of feature-values.
http://wals.info/combinations/24A_58A_58B#2/25.5/157.5 has 26 zeroes and 14 non-zeroes.
http://wals.info/combinations/59A_24A_58B#2/25.5/157.5 has 57 zeroes and 23 non-zeroes.
I may have gotten most of that last paragraph wrong. But you can see how to find out for yourself.
Edit: Recounted; pretty sure I got them right this time.
BTW; WALS.info now lets you sort the combinations by number of languages.