results$dataset is inconsistently matrix, or df

[maxheld83]

... depending on the input to qmethod.

Check this out:

str(lipset[[1]])
str(invisible(qmethod(dataset = lipset[[1]], nfactors = 3, forced = TRUE))$dataset)
str(as.matrix(lipset[[1]]))
str(invisible(qmethod(dataset = as.matrix(lipset[[1]]), nfactors = 3, forced = TRUE))$dataset)

yields:

> str(lipset[[1]])
'data.frame':   33 obs. of  9 variables:
 $ US1: int  -1 0 -2 0 -2 1 0 -1 0 -1 ...
 $ US2: int  -1 0 -1 -3 2 3 1 1 -4 0 ...
 $ US3: int  2 -2 -2 4 -1 0 -4 -3 1 -4 ...
 $ US4: int  3 1 -3 -1 -1 3 -3 -2 0 -4 ...
 $ JP5: int  -4 -1 3 -1 1 1 4 2 -4 4 ...
 $ CA6: int  1 -3 0 3 3 4 -2 0 -2 -2 ...
 $ UK7: int  2 0 -2 1 0 1 -1 -3 0 -1 ...
 $ US8: int  -2 2 0 -3 -4 4 0 -1 -1 -1 ...
 $ FR9: int  3 1 0 1 -4 -3 2 2 -2 0 ...
> str(invisible(qmethod(dataset = lipset[[1]], nfactors = 3, forced = TRUE))$dataset)
Q-method analysis.
Finished on:             Fri Aug 21 19:32:14 2015
Original data:           33 statements, 9 Q-sorts
Forced distribution:     TRUE
Number of factors:       3
Rotation:                varimax
Flagging:                automatic
Correlation coefficient: pearson
'data.frame':   33 obs. of  9 variables:
 $ US1: int  -1 0 -2 0 -2 1 0 -1 0 -1 ...
 $ US2: int  -1 0 -1 -3 2 3 1 1 -4 0 ...
 $ US3: int  2 -2 -2 4 -1 0 -4 -3 1 -4 ...
 $ US4: int  3 1 -3 -1 -1 3 -3 -2 0 -4 ...
 $ JP5: int  -4 -1 3 -1 1 1 4 2 -4 4 ...
 $ CA6: int  1 -3 0 3 3 4 -2 0 -2 -2 ...
 $ UK7: int  2 0 -2 1 0 1 -1 -3 0 -1 ...
 $ US8: int  -2 2 0 -3 -4 4 0 -1 -1 -1 ...
 $ FR9: int  3 1 0 1 -4 -3 2 2 -2 0 ...
> str(as.matrix(lipset[[1]]))
 int [1:33, 1:9] -1 0 -2 0 -2 1 0 -1 0 -1 ...
 - attr(*, "dimnames")=List of 2
  ..$ : chr [1:33] "sta_1" "sta_2" "sta_3" "sta_4" ...
  ..$ : chr [1:9] "US1" "US2" "US3" "US4" ...
> str(invisible(qmethod(dataset = as.matrix(lipset[[1]]), nfactors = 3, forced = TRUE))$dataset)
Q-method analysis.
Finished on:             Fri Aug 21 19:32:14 2015
Original data:           33 statements, 9 Q-sorts
Forced distribution:     TRUE
Number of factors:       3
Rotation:                varimax
Flagging:                automatic
Correlation coefficient: pearson
 int [1:33, 1:9] -1 0 -2 0 -2 1 0 -1 0 -1 ...
 - attr(*, "dimnames")=List of 2
  ..$ : chr [1:33] "sta_1" "sta_2" "sta_3" "sta_4" ...
  ..$ : chr [1:9] "US1" "US2" "US3" "US4" ...

This can cause unexpected problems (it was the root cause of kernel panic in #236).

Maybe we want to streamline this at some point, so that results$dataset is always either a data.frame or a matrix.

I'm in favor of the matrix, because all the cells are, in fact, of the same data type (integers).

[aiorazabala]

This depends on whether the initial (raw) data is a matrix or a data.frame. I agree that we can just add a code line at the beginning of qmethod() to transform whatever comes into a matrix.