Using -asdoc- to get summary statistics and weights

[Lorien Nair]

Hi Everyone,

I am using the 2015-16 wave of the Demographic and health survey from India, it contains data on reproductive behaviour and contraception. I have produced summary statistics using asdoc by Attaullah Shah, however I wanted to get summary statistics that would show proportions of a category variable. For example if I wanted to get the summary statistics of say proportion of people belonging to a certain category, where the category could be as follows:

 label list `:value label occupation '
V717:
           0 not in work force/no occupation
           1 professional/technical/managerial
           2 clerical
           3 sales
           5 agricultural
           7 services/household and domestic
           8 manual - skilled and unskilled
          98 don't know

label list `:value label state'
1 Andhra Pradesh
2 Assam 
3 Bihar 
.
.

I want to be able to run asdoc to get the proportion of people belonging to each category to be a part of my summary statistics since in this scenario having the mean of a categorical variable in a survey of individuals does not make much sense.

Also, does asdoc incorporate weights while generating tables, I have not found any material alluding to that.

Thanks in advance.

Lori

[Attaullah Shah]

Please use dataex (from SSC) and post an example dataset here. Seems to me you may want to use the tabulate command with asdoc.
see

help asdoc

and browse to the Section 5 of the help file.

[Lorien Nair]

Hi Dr. Shah,

I do not need to tabulate the data instead I wanted to include percentage of each category in my summary statistics.

For example for occupation in the summary stats I want a breakdown of what percentage of the respondents belong to each category.

* Example generated by -dataex-. To install: ssc install dataex
clear
input byte occupation
 .

 .
 .
 0
 8
 .
 0
 0
 0
 0
 0
 .
 0
 .
98
 0
 .
 .
 .
 .
 2
 1
 0
 .

 
end
label values occupation V717
label def V717 0 "not in work force/no occupation", modify
label def V717 1 "professional/technical/managerial", modify
label def V717 2 "clerical", modify
label def V717 8 "manual - skilled and unskilled", modify
label def V717 98 "don't know", modify

I have attached an example from a published paper. Thank you so much.

Lori

[Attaullah Shah]

The dataset you have posted is not usable as it is truncated. The picture you have posted for the desired table shows that the percentage variable is actually a mean of something. Therefore, you can get it by using the stat() option of asdoc. see this example

webuse grunfeld
asdoc sum kstock mvalue, stat(N mean sd median)

[Lorien Nair]

Thank you for your guidance Prof. Shah.

But when I run the asdoc sum command, I get the following output:

asdoc by year: sum age religion educ_single educ_lvl educ_att  tot_child_born currently_preg wanted_currpreg abort_ever contra_current contra_currenttype use_pattern child_wantlast ///
marital_status_current marital_status unmet_need part_educ_year part_occgroup emp_home reading_freq hh_radio hh_tele radio_use tv_use fp_radio fp_tv, stat(mean median sd N) replace label dec(2)

2005 - NFHS III
Variable	N	Mean	Sd	Min	Max
Respondent's current age	7947	28.817	9.487	15	49
Religion	7947	1.502	3.498	1	99
Education in single years	7947	5.499	5.289	0	99
Highest educational level	7947	1.21	1.027	0	9
Educational attainment	7947	1.897	1.682	0	9
Total children ever born	7947	2.244	2.14	0	12
Currently pregnant	7947	0.046	0.21	0	1
Current pregnancy wanted	368	1.446	0.674	1	3
Ever had a terminated pregnancy	7947	0.153	0.372	0	9
Current contraceptive method	7947	2.556	3.188	0	18
Current use by method type	7947	1.258	1.437	0	3
Pattern of use	7947	2.532	1.432	1	4
Wanted last child	2403	1.438	0.746	1	9
Current marital status (see s301	7947	0.871	0.706	0	5
Currently/formerly/never in unio	7947	0.81	0.484	0	2
Unmet need	7947	3.947	5.447	0	99
Husband/partner's highest year o	6155	4.251	10.585	0	99
Husband/partner's occupation (gr	6155	5.799	6.657	0	99
Respondent works for family, oth	2852	1.704	0.735	1	3
Frequency of reading newspaper o	7947	0.793	1.125	0	9
Household has: radio	7947	0.676	1.519	0	9
Household has: television	7947	0.918	1.458	0	7
Frequency of listening to radio	7947	0.936	1.188	0	9
Frequency of watching television	7947	1.865	1.287	0	9
Heard family planning on radio l	7947	0.334	0.518	0	9
Heard family planning on tv last	7947	0.573	0.53	0	9

As you can see, the categorical variables do not make sense or convey any information. Would it be more sensible if I were to use categorical variables like religion or household has radio , then should I manually tab the variable with year and then fill out the table. Is there a way to get this output in another way?

Best,

Lori

[Lorien Nair]

Since I wanted the percentage of people belonging to each category in a categorical variable, I was able to generate the table below. However, I had to manually tab out the categorical variable and then insert them into the table after running asdoc sum. Is there an easier way to do this?

Descriptive Statistics
year of interview: 1992
Variable	Mean (/%)	Median	sd	N
respondent's current age	30.64	30	8.93	37142
religion	2.38	1	10.65	36988
Hindu (%)	84.97	.	.	31,427
Muslim (%)	10.47	.	.	3,871
Christian (%)	3.29	.	.	1,218
Other (%)	1.28	.	.	472
education in single years	3.37	0	6.51	37142
highest educational level	0.71	0	.	37142
no education (%)	57.9	.	.	21,504
primary (%)	18.93	.	.	7,030
secondary (%)	19.46	.	.	7,227
higher (%)	3.46	.	.	1,285
total children ever born	3.02	3	2.24	37142
currently pregnant	0.08	0	.	37142
no or unsure (%)	92.17	.	.	34,232
yes (%)	7.83	.	.	2,910
household has: radio	0.46			37142
no (%)	54.71			20,319
yes (%)	45.2			16,787
household has: television	0.24			37142
no (%)	77.01			28,604
yes (%)	22.91			8,508
heard family planning on radio l	0.38			37142
no (%)	62.64			23,264
yes (%)	37.34			13,867
heard family planning on tv last	0.3			37142
no (%)	70.76			26,280
yes (%)	29.19			10,840

Thanks a lot!

Lori

[Attaullah Shah]

It is hard to figure out a solution here without seeing the actual data and the way you are calculating the percentage. As suggested in #2, post a data example that can be used on Stata and the code that you are using for finding the percentages.

[Lorien Nair]

Dear Prof. Shah,

I use the -tab- command to get percentages since they are categorical variables, so:

tab religion if year == 1992

dataex age religion educ_lvl currently_preg

* Example generated by -dataex-. To install: ssc install dataex

clear
input byte age float religion byte(educ_lvl currently_preg)
46 3 0 0
20 3 3 0
40 1 1 0
21 1 3 0
19 1 3 0
26 1 2 0
43 1 2 0
17 1 2 0
22 1 2 0
35 1 2 0
25 1 3 0
46 3 2 0
17 2 2 0
40 3 2 0
21 3 2 0
17 3 2 0
46 1 3 0
40 1 2 0
37 1 2 0
37 1 2 0
33 3 2 0
44 1 2 0
24 1 3 0
21 1 3 0
42 1 3 0
35 1 3 0
27 1 3 0
34 2 3 0
20 3 2 0
24 3 2 0
40 3 1 0
33 3 2 0
32 3 2 0
47 3 1 0
19 3 2 0
48 3 2 0
24 3 2 0
32 3 2 0
20 3 2 0
30 3 2 0
38 3 2 0
31 3 2 0
40 3 2 0
22 3 2 0
43 3 2 0
23 3 2 0
17 3 2 0
20 3 2 0
45 3 1 0
15 3 2 0
31 3 2 0
46 3 0 0
29 3 2 0
41 3 2 0
40 3 2 0
18 3 0 0
16 3 2 0
27 3 2 0
16 3 2 0
30 3 0 0
40 3 1 0
28 3 2 0
38 3 1 0
44 3 0 0
32 3 2 0
43 2 2 0
17 2 2 0
49 3 2 0
31 3 2 0
42 3 1 0
46 3 2 0
29 3 2 0
37 3 2 0
36 3 2 0
48 3 2 0
22 3 2 0
49 3 2 0
40 3 2 0
29 3 2 0
32 3 2 0
48 3 1 0
18 3 2 0
29 3 2 0
25 3 2 0
43 3 2 0
17 3 2 0
15 3 2 0
37 3 2 0
41 3 3 0
39 3 1 0
39 3 1 0
46 3 2 0
43 3 0 0
39 3 2 0
39 3 2 0
39 3 2 0
19 3 2 0
45 3 2 0
22 3 2 0
36 3 2 0
end
label values religion V130
label def V130 1 "hindu", modify
label def V130 2 "muslim", modify
label def V130 3 "christian", modify
label values educ_lvl V106
label def V106 0 "no education", modify
label def V106 1 "primary", modify
label def V106 2 "secondary", modify
label def V106 3 "higher", modify
label values currently_preg V213
label def V213 0 "no or unsure", modify