Boxplots

[Joliek44]

Hello, 

I am trying to figure out how to boxplot the information below. I have been using the below equation and many other variations in SAS and has not been working. Not sure its possible, but would like assistance on trying to boxplot the data. Any suggestions would be great. 

proc boxplot data=work.wtloss;

plot pre*post wtloss;

run;

pat_id	pre_weight	post_weight
1	165	160
2	202	200
3	256	259
4	155	156
5	135	134
6	175	162
7	180	187
8	174	172
9	136	138
10	168	162
11	207	197
12	155	155
13	220	205
14	163	153
15	159	150
16	253	255
17	138	128
18	287	280
19	177	171
20	181	170
21	148	154
22	167	170
23	190	180
24	165	154
25	155	150
26	153	145
27	205	206
28	186	184
29	178	166
30	129	132
31	125	127
32	165	169

[PaigeMiller · Accepted solution]

Your data set is not structured properly to permit boxplots to be drawn by PROC BOXPLOT.

data wtloss1;
    set wtloss;
    length time $ 4;
    time='PRE';
    wt=pre_weight;
    output;
    time='POST';  
    wt=post_weight;
    output;
    drop pre_weight post_weight;
run;
proc boxplot data=wtloss1;
    plot wt*time;
run;

Having said that, I think this is the WRONG boxplot to analyze the data. This seems to be paired measurements of weight, and the true quantity of interest is the difference between PRE and POST. In that case, you want to compute the difference in a DATA step, and then create a boxplot of the difference.

[Joliek44]

Hello Paige, 

So, do you think its better to just boxplot each group separately, by pat id and pre weight and pat id and post weight?  

[PaigeMiller]

---

@Joliek44 wrote:

Hello Paige, 

So, do you think its better to just boxplot each group separately, by pat id and pre weight and pat id and post weight?  

---

No, this is not my recommendation. My recommendation is to create a boxplot of the differences. I don't think boxplot of PRE and boxplot of POST (in any manner of display) is not helpful here. But that's my understanding, an educated guess really, because you really haven't stated why you are analyzing this data and what you want to get out of the analysis.

[Joliek44]

Yes, for sure. I did run the data and got boxplot done. Thank you.