I am using -corr2data- to simulated raw data from a correlation matrix. However, some variables that I need should be binary. How can I convert?
Is it possible to convert higher amounts to 1 (and the other ones to 0) as the form to reach the same mean? How should I do it?
Is a way in R?
(I want to perform a GSEM on a correlation matrix)
(I know -faux- package in R. But my problem is that just some of [not all of] my variables are binary.)
perhaps this
https://www.stata.com/statalist/archive/2003-07/msg00054.html
might help somehow
Steftcho P. Dokov
Thank you for your suggestion
As I found, you introduced a command to create a random binary variable with a known mean
However, I want to convert a continuous variable to a binary variable with same mean
For example:
corr2data x y, n(1000) means(0.5 0.5) sds(0.5 0.5) corr(C)
x, y --> binary ???
I'm not sure that this is possible. A normal is continuous while a Bernoulli is discreet. You might at the attached.perhaps you mean something else. David Booth
Steftcho P. Dokov
Thanks! What was your command? Does it work for non-zero correlations?
David Eugene Booth
As you simulate raw data from correlation matrix using -faux- package, you can not choose some variables as continuous and some others as binary. you have to choose norm() for all (we can type same mean [p] and sd [√pq], but we receive a continuous result) and then convert them to binary.
Please see this help. I am looking for something like norm2bernoulli!!!
https://cran.r-project.org/web/packages/faux/faux.pdf
I hope you can see the MS Excel built-in functions that are utilized in the columns of the attachment file.
I found a way in stata:
x ---> is the simulated variable with mean "e.g. 0.3".
_pctile x, p(70)
di r(r1)
the result of p70th will be shown = ???
generate x_binary = 1 if x > ???
- Badges
- Science topic
- Similar topics
- Geoscience
- Cartography
- Mapping
- Maps