statsample-glm
statsample-glm copied to clipboard
#coefficients output type
I want to talk about two things:
(1). #coefficients(:hash) return vector names mapping the coefficients but the type of vector names is not preserved. They are all converted to symbols. This leads to problem in prediction as the new_data has to contain all vector names as symbols.
[6] pry(main)> x.coefficients :hash
=> {:a=>1.144620819921843,
:b=>-0.042921191993683674,
:"a:b"=>-0.03010679572782158,
:constant=>4.738219960853547}
(2). #coefficients(:vector) returns a vector as follow:
=> #<Daru::Vector(4)>
0 1.144620819921843
1 -0.04292119199368367
2 -0.03010679572782158
3 4.738219960853547
The vector is not indexed by vector names which is little inconvenient.
WDYT about these two?
(1). #coefficients(:hash) return vector names mapping the coefficients but the type of vector names is not preserved. They are all converted to symbols. This leads to problem in prediction as the new_data has to contain all vector names as symbols.
The keys should be of the same type as the names of the vector, i.e. symbol if it's originally a symbol and something else otherwise.
(2). #coefficients(:vector) returns a vector as follow:
Yes this should be indexed on the vector names.