I have a dataset in which I want to understand correlations across several variables. Within this data set are numerous user groups in which this analysis needs to be run and I am trying to do so without creating a bunch of data subsets. Is this possible? So far selecting the appropriate cases and running the analysis does so for all cases rather than what is selected.
I thought about that and I think it would work in cases where I have people predefined into a single variable. My concern is when that is not the case, that is I have not created variable for a group based but want to do something more on the fly.
Also, since you brought up "By", let's say I have a set of data collected among a rep sample of Widget A users. Widget A users were asked to test 1 of 5 different Gadgets. I want to run a similar analysis By Gadget but only among Widget A users that are Male, 18-24 and live in the Northwest.
It seems like just using By right off the bat is not going to work as it would give me the correlations by...
and, assuming I have a variable where this group is collected
Does that make sense and am I understanding the capabilities and limitations of "By"?
I have two suggestions:
1. Create a new column that groups data the way you want it. Use this new column as your BY variable. This might be difficult if you have many different ways you want to slice the data
2. Select the desired rows (CTRL-Shift-W brings up SELECT WHERE). Invert the selection and exclude the rows. This is easier to do programmatically if you are handy with JSL.
dt = open("$sample_data\Car Physical Data.jmp");
dt << select where(:Country == "USA" & :Type == "Sporty" & :Horsepower > 150)
<< invert row selection << exclude << clear select;
// Analysis goes here...