topic Re: PROC CORRESP MCA with 'Yes/No' data in Statistical Procedures

PROC CORRESP MCA with 'Yes/No' data

DR500 — Sat, 25 Apr 2020 18:41:03 GMT

Hi All,

Need a little guidance here. I have categorical Yes/No data for multiple fields. I'm looking to cluster this using the MCA option in PROC CORRESP. However, when SAS spits out my Burt Table and my graph, it is not identifying which field the Yes or No answer belongs to. For instance, if my Field Name was 'HasBankAccount' with Yes or No as the answer, the Burt table will show 'Yes' and 'No', but does not say what field each 'Yes' and 'No' belongs to. I want it to show 'HasBankAccount_Yes' and 'HasBankAccount_No' or something similar just so I know which fields I'm looking at in my cluster.

Anyone have any ideas? I've looked all over the internets and got nothing.

Thanks in advance! - Dave

Re: PROC CORRESP MCA with 'Yes/No' data

PaigeMiller — Sat, 25 Apr 2020 20:17:20 GMT

Show us a representative portion of the data. Show us the output.

Re: PROC CORRESP MCA with 'Yes/No' data

DR500 — Sat, 25 Apr 2020 20:32:18 GMT

Ok, here's a rudimentary mock-up:

Dataset
CustID	HasAccount	HasSavings
1	Yes	No
2	Yes	Yes
3	No	No

Burt Table

	Yes	No	Yes	No
Yes	1	1	1	1
No	2	2	2	2
Yes	1	1	1	1
No	2	2	2	2

Desired Output

	HasAccount_Yes	HasAccount_No	HasSavings_Yes	HasSavings_No
HasAccount_Yes	1	1	1	1
HasAccount_No	2	2	2	2
HasSavings_Yes	1	1	1	1
HasSavings_No	2	2	2	2

R seems to be able to do this with FactomineR, but I really want to do this in SAS.

Does this make more sense? Ignore the actual counts - I just made up the ones and twos.

Thanks!

Dave

Re: PROC CORRESP MCA with 'Yes/No' data

PaigeMiller — Sat, 25 Apr 2020 21:26:56 GMT

If you look at this example, the levels of each variable are formatted differently, and so the Burt Table is interpretable. They accomplish this (as you should) by assigning formats to the original data, so then you can separate the Yes under "HasAccount" from the Yes under "Has Savings"

Re: PROC CORRESP MCA with 'Yes/No' data

DR500 — Sat, 25 Apr 2020 23:28:15 GMT

Thanks. I was hoping there was an easier solution than formatting each field(I actually have a lot of fields that would need formatting). Looks like I'm switching to R. Appreciate the help.

- Dave

Re: PROC CORRESP MCA with 'Yes/No' data

Ksharp — Mon, 27 Apr 2020 10:54:54 GMT

Why not change your data?

P.S. Fixed it now, Should test it firstly .

data want;
set have;
array x{*} $ _character_;
do i=1 to dim(x);
x{i}=catx('_',vname(x{i}),x{i});
end;
run;

Re: PROC CORRESP MCA with 'Yes/No' data

PaigeMiller — Sun, 26 Apr 2020 11:32:12 GMT

Great idea, @Ksharp

Hey, @DR500 you should give this a try.

Re: PROC CORRESP MCA with 'Yes/No' data

DR500 — Sun, 26 Apr 2020 18:02:56 GMT

Thanks! I tried what you posted and received the following error:

Array subscript out of range at line 75 column 21.

Is this the result of a length issue? If so, is there an easy way to correct it? I'd like to do it with script, but if necessary, I can brute force it by writing formulas in Excel to modify. FYI - extensive SQL background but fairly new to SAS, so please bear with me on this! Thanks! - Dave

Re: PROC CORRESP MCA with 'Yes/No' data

PaigeMiller — Sun, 26 Apr 2020 18:04:51 GMT

Show us the log (not just the error message, but the entire log so we can see the code you used and the NOTEs and WARNINGs and ERRORs, unedited, verbatim)

Re: PROC CORRESP MCA with 'Yes/No' data

DR500 — Sun, 26 Apr 2020 18:27:00 GMT

1 OPTIONS NONOTES NOSTIMER NOSOURCE NOSYNTAXCHECK;

71 data work.test;

72 set work.custdata_mod;

73 array x{*} $ _character_;

74 do _n_=1 to dim(x);

75 x{i}=catx('_',vname(x{i}),x{i});

76 end;

77 run;

NOTE: Variable i is uninitialized.

ERROR: Array subscript out of range at line 75 column 21.

customerID=7590-VHVEG gender=Female Partner=Yes Dependents=No tenure=1 PhoneService=No MultipleLines=No InternetService=DSL

OnlineSecurity=No OnlineBackup=Yes DeviceProtection=No TechSupport=No StreamingTV=No StreamingMovies=No Contract=Month-to-month

PaperlessBilling=Yes PaymentMethod=Electronic check MonthlyCharges=29.85 TotalCharges=29.85 SeniorCitizen=No Churn=0

Tenure_Group=0-12 Months MonthlyCharges_Group=0-35 DollarsPerMonth i=. _ERROR_=1 _N_=1

NOTE: The SAS System stopped processing this step because of errors.

NOTE: There were 1 observations read from the data set WORK.CUSTDATA_MOD.

WARNING: The data set WORK.TEST may be incomplete. When this step was stopped there were 0 observations and 24 variables.

NOTE: DATA statement used (Total process time):

real time 0.00 seconds

user cpu time 0.01 seconds

system cpu time 0.00 seconds

memory 2450.03k

OS Memory 36268.00k

Timestamp 04/26/2020 05:34:03 PM

Step Count 70 Switch Count 2

Page Faults 0

Page Reclaims 433

Page Swaps 0

Voluntary Context Switches 9

Involuntary Context Switches 0

Block Input Operations 0

Block Output Operations 264

79 OPTIONS NONOTES NOSTIMER NOSOURCE NOSYNTAXCHECK;

Re: PROC CORRESP MCA with 'Yes/No' data

PaigeMiller — Sun, 26 Apr 2020 20:03:09 GMT

Seems like both Ksharp and I missed the obvious mistake.

You want to use

do i=1 to dim(x);

Re: PROC CORRESP MCA with 'Yes/No' data

DR500 — Sun, 26 Apr 2020 20:20:30 GMT

That fixed it! However, it modified all the character fields (which is what the script told it to do) but instead of using _character_, can I list only the fields(variables) that I want changed? Like:

array x{*} $ OnlineSecurity OnlineBackup DeviceProtection;

Or create a macro and call it for the fields I need like:

array x{*} $ &mod_fields;

Would either of those work? Thanks for all the help on this!

Re: PROC CORRESP MCA with 'Yes/No' data

DR500 — Sun, 26 Apr 2020 20:37:39 GMT

Nevermind - I just played around with it and it ran fine with a macro. Really appreciate all the help!

Re: PROC CORRESP MCA with 'Yes/No' data

PaigeMiller — Sun, 26 Apr 2020 20:39:55 GMT

Sure that works, but let's get the terminology correct.

&mod_fields is a macro variable, it is not a macro. Macro variables are different than macros, and hopefully using the correct name can avoid confusion in the future.

Re: PROC CORRESP MCA with 'Yes/No' data

DR500 — Sun, 26 Apr 2020 21:16:47 GMT

Got it. Like I said, coming from 15+ years of SQL to this is interesting - all sorts of different terminology. Thanks again!