Creating a group variable from several dummy indicators

Shad — Wed, 20 Mar 2019 17:55:14 GMT

Hello,

I'm trying to create observational groups based on the types of post operative pain medications that a patient received. I have 5 dummy variables coded 1/0, and have been trying to use those 5 variables to account for all combinations and create a variable that contains each combo.

However something isn't quite working right. Hopefully it's something simple with the way that I'm using IF/THEN statements.

To account for all combinations of my dummy variables I did a multidimensional contingency table.

proc freq data = ty_analysis;
table pre_op*DOS_op*post_op pre_ty*DOS_ty*post_ty pre_keto*DOS_keto*post_keto pre_ibu*DOS_ibu*post_ibu pre_nar*DOS_nar*post_nar /list; 
run;

post_op	post_IV	post_ty	post_keto	post_ibu	Frequency	Percent	Cumulative Frequency	Cumulative Percent
0	0	0	0	0	23711	20.16	23711	20.16
0	0	0	0	1	1281	1.09	24992	21.25
0	0	0	1	0	10378	8.82	35370	30.08
0	0	0	1	1	592	0.50	35962	30.58
0	0	1	0	0	3497	2.97	39459	33.55
0	0	1	0	1	1737	1.48	41196	35.03
0	0	1	1	0	5998	5.10	47194	40.13
0	0	1	1	1	863	0.73	48057	40.86
0	1	1	0	0	243	0.21	48300	41.07
0	1	1	0	1	96	0.08	48396	41.15
0	1	1	1	0	987	0.84	49383	41.99
0	1	1	1	1	248	0.21	49631	42.20
1	0	0	0	0	18730	15.93	68361	58.13
1	0	0	0	1	2854	2.43	71215	60.55
1	0	0	1	0	17927	15.24	89142	75.80
1	0	0	1	1	1606	1.37	90748	77.16
1	0	1	0	0	6300	5.36	97048	82.52
1	0	1	0	1	2347	2.00	99395	84.52
1	0	1	1	0	9295	7.90	108690	92.42
1	0	1	1	1	2940	2.50	111630	94.92
1	1	1	0	0	1130	0.96	112760	95.88
1	1	1	0	1	430	0.37	113190	96.25
1	1	1	1	0	3034	2.58	116224	98.83
1	1	1	1	1	1381	1.17	117605	100.00

Using the the table to account for all groups I coded my postgroup variable

if post_op EQ 1 & post_IV EQ 1 & post_ty EQ 1 & post_keto EQ 1 & post_ibu EQ 1 then postgroup = " OTAKI ";
else if post_op EQ 1 & post_IV EQ 1 & post_ty EQ 1 & post_keto EQ 1 & post_ibu EQ 0 then postgroup = " OTAK ";
else if post_op EQ 1 & post_IV EQ 1 & post_ty EQ 1 & post_keto EQ 0 & post_ibu EQ 1 then postgroup = " OTAI ";
else if post_op EQ 1 & post_IV EQ 1 & post_ty EQ 1 & post_keto EQ 0 & post_ibu EQ 0 then postgroup = " OTA ";
else if post_op EQ 1 & post_IV EQ 0 & post_ty EQ 1 & post_keto EQ 1 & post_ibu EQ 1 then postgroup = " OAKI ";
else if post_op EQ 1 & post_IV EQ 0 & post_ty EQ 1 & post_keto EQ 1 & post_ibu EQ 0 then postgroup = " OAK ";
else if post_op EQ 1 & post_IV EQ 0 & post_ty EQ 1 & post_keto EQ 0 & post_ibu EQ 1 then postgroup = " OAI ";
else if post_op EQ 1 & post_IV EQ 0 & post_ty EQ 1 & post_keto EQ 0 & post_ibu EQ 0 then postgroup = " OA ";
else if post_op EQ 1 & post_IV EQ 0 & post_ty EQ 0 & post_keto EQ 1 & post_ibu EQ 1 then postgroup = " OKI ";
else if post_op EQ 1 & post_IV EQ 0 & post_ty EQ 0 & post_keto EQ 1 & post_ibu EQ 0 then postgroup = " OK ";
else if post_op EQ 1 & post_IV EQ 0 & post_ty EQ 0 & post_keto EQ 0 & post_ibu EQ 1 then postgroup = " OI ";
else if post_op EQ 1 & post_IV EQ 0 & post_ty EQ 0 & post_keto EQ 0 & post_ibu EQ 0 then postgroup = " O ";
else if post_op EQ 0 & post_IV EQ 1 & post_ty EQ 1 & post_keto EQ 1 & post_ibu EQ 1 then postgroup = " TAKI ";
else if post_op EQ 0 & post_IV EQ 1 & post_ty EQ 1 & post_keto EQ 1 & post_ibu EQ 0 then postgroup = " TAK ";
else if post_op EQ 0 & post_IV EQ 1 & post_ty EQ 1 & post_keto EQ 0 & post_ibu EQ 1 then postgroup = " TAI ";
else if post_op EQ 0 & post_IV EQ 1 & post_ty EQ 1 & post_keto EQ 0 & post_ibu EQ 0 then postgroup = " TA ";
else if post_op EQ 0 & post_IV EQ 0 & post_ty EQ 1 & post_keto EQ 1 & post_ibu EQ 1 then postgroup = " AKI ";
else if post_op EQ 0 & post_IV EQ 0 & post_ty EQ 1 & post_keto EQ 1 & post_ibu EQ 0 then postgroup = " AK ";
else if post_op EQ 0 & post_IV EQ 0 & post_ty EQ 1 & post_keto EQ 0 & post_ibu EQ 1 then postgroup = " AI ";
else if post_op EQ 0 & post_IV EQ 0 & post_ty EQ 1 & post_keto EQ 0 & post_ibu EQ 0 then postgroup = " A ";
else if post_op EQ 0 & post_IV EQ 0 & post_ty EQ 0 & post_keto EQ 1 & post_ibu EQ 1 then postgroup = " KI ";
else if post_op EQ 0 & post_IV EQ 0 & post_ty EQ 0 & post_keto EQ 1 & post_ibu EQ 0 then postgroup = " K ";
else if post_op EQ 0 & post_IV EQ 0 & post_ty EQ 0 & post_keto EQ 0 & post_ibu EQ 1 then postgroup = " I ";
else if post_op EQ 0 & post_IV EQ 0 & post_ty EQ 0 & post_keto EQ 0 & post_ibu EQ 0 then postgroup = " NONE ";

The problem is that running a proc freq on postgroup, the counts for each grouping don't add up and I have some obs for instance that are 0 in all dummy variables being added to different groups instead of the NONE category.

postgroup	Frequency	Percent	Cumulative Frequency	Cumulative Percent
NONE	22547	19.17	22547	19.17
O	19020	16.17	41567	35.34
OK	18380	15.63	59947	50.97
K	10513	8.94	70460	59.91
OAK	9343	7.94	79803	67.86
OA	6327	5.38	86130	73.24
AK	6018	5.12	92148	78.35
A	3510	2.98	95658	81.34
OTAK	3034	2.58	98692	83.92
OAKI	2952	2.51	101644	86.43
OI	2939	2.50	104583	88.93
OAI	2360	2.01	106943	90.93
AI	1750	1.49	108693	92.42
OKI	1633	1.39	110326	93.81
OTAKI	1381	1.17	111707	94.98
I	1301	1.11	113008	96.09
OTA	1131	0.96	114139	97.05
TAK	987	0.84	115126	97.89
AKI	868	0.74	115994	98.63
KI	594	0.51	116588	99.14
OTAI	430	0.37	117018	99.50
TAKI	248	0.21	117266	99.71
TA	243	0.21	117509	99.92
TAI	96	0.08	117605	100.00

post_op	post_IV	post_ty	post_keto	post_ibu	postgroup	Frequency	Percent	Cumulative Frequency	Cumulative Percent
0	0	0	0	0	KI	2	0.34	2	0.34
0	0	0	1	1	KI	592	99.66	594	100.00

Re: Creating a group variable from several dummy indicators

ballardw — Wed, 20 Mar 2019 16:46:20 GMT

Without your data and complete code it is almost impossible to determine why as we would need examples that create the incorrect assignment.

I am going to suggest an alternate form of coding the postgroup that might help as that many if then else statements have lots of potential for headaches.

proc format library=work;
value O 1='O' other=' ';
value T 1='T' other=' ';
value A 1='A' other=' ';
value K 1='K' other=' ';
value I 1='I' other=' ';
RUN;


data example;
   input post_op post_iv post_ty post_keto post_ibu;
   length postgroup $ 5. ;
   postgroup= cats(put(post_op,o.),put(post_iv, t.),put(post_ty, a.),put(post_keto, k.),put(post_ibu, i.) );
   if missing(postgroup) then postgroup='NONE';
datalines;
0 0 0 0 0
0 1 0 0 1
1 1 1 1 1
1 0 1 0 1
;
run;

I would suggest that the most likely issue relates to how you get the proc freq output into a single data set as your example "data" does not match the output that would be created by the proc freq code you show.

topic Creating a group variable from several dummy indicators in SAS Programming

Creating a group variable from several dummy indicators

Re: Creating a group variable from several dummy indicators

Re: Creating a group variable from several dummy indicators