topic Re: proc factor number of observations used in Statistical Procedures

proc factor number of observations used

Golumn — Mon, 18 Feb 2019 21:10:38 GMT

Hi, I was on here a little bit ago for proc factor vs. proc pls.

In the case of the dataset I currently have I've been advised to continue using proc factor.

I've done alot of work in ensuring my dataset has numeric variables, and using dummy variables if required.

I then proceeded with below:

proc factor data=dataset_used

method = principle mineigen=0

rotate=varimax reorder

outstat=all_pattern;

run;

my initial intention was to not limit via factor=# and see how many factors have an eigen >=1.

then re-run while applying a factor limitation.

When i looked at the output i noticed the number of records used was different then the number of records observed....

Question1: what causes the difference in records used? is it the presence of null values within at least one variable in an observational row?

Question2: if I don't specify a variable list, does it default to using all variables?

Question3: And if you see anything stupid please feel free to tell me.

Re: proc factor number of observations used

PaigeMiller — Mon, 18 Feb 2019 21:32:27 GMT

@Golumn wrote:

Question1: what causes the difference in records used? is it the presence of null values within at least one variable in an observational row?

Question2: if I don't specify a variable list, does it default to using all variables?

Question3: And if you see anything stupid please feel free to tell me.

Answer 1: yes, missing values is the cause

Answer 2: probably it will use all numeric variables, but I never tried

Answer 3: just pointing out that if you use PROC PLS, you don't have to create the dummy variables first, you can use the CLASS statement and then the procedure will internally create the dummy variables

Re: proc factor number of observations used

Golumn — Mon, 18 Feb 2019 21:35:05 GMT

Thanks again for the feedback!