FW: Survey sampling weight calculation

[John F Hall]

Malc, Georgie,

 

There’s been a series of exchanges on the Survey Research Methods Section of the American Statistical Association list concerning how to weight responses from a survey in Iran.  The survey seems to have been done in a professional way, but the researcher was apparently not involved in the design and data collection stages and was asking the list members for help in weighting the data using Stata.  Earlier exchanges were very technical, but this latest contribution raises some important ethical questions which should interest you, your students and members of your professional associations.

 

John

 

 

John F Hall (Mr)

[Retired academic survey researcher]

 

Email:   [hidden email] 

Website: www.surveyresearch.weebly.com

SPSS start page:  www.surveyresearch.weebly.com/spss-without-tears.html

  

  

 

 

 

 

 

From: Survey Research Methods Section of the ASA [mailto:[hidden email]] On Behalf Of David Judkins
Sent: 25 June 2013 02:50
To: [hidden email]
Subject: Re: Survey sampling weight calculation

 

Amir is an interesting case.  His heart seems to be in the right place.  He wants to do something rigorous with his data.  He is knows next to nothing about sampling statistics, but he is bold and friendly enough to reach out for help.  Technically, I think he is an employee of the government of Iran.  This means that providing him technical assistance may violate U.S. laws.  It certainly violates the spirit of the trade embargo.  We in the U.S. want life to be hard in Iran until they agree to stop enriching uranium. From my perspective, it would also be super if they would not send weapons to Hezbollah and both weapons and soldiers to support Assad in Syria. Better yet if they would apologize for the hostage taking in 1979 and pay reparations to their victims.  At one point, I became a friend of one of the former hostages.  He underwent two mock executions.  He and the other surviving former hostages are still waiting for justice.  The reason why Amir cannot afford to buy basic textbooks is that he lives under a theocracy that is hostile to us (the Great Satan).  On the other hand, why not reach out the olive branch?  Amir is taking some personal risks here himself, asking westerners for help on an internal Iranian survey.  So shall we help him or not? 

 

Amir has told us nothing about his study.  What is the content?  If it were about how to build support for polio vaccination, I would be happy to help him. Obviously, if it is about evaluating recruiting strategies for volunteers to support Iran's covert intervention in Syria, then I would have nothing to do with it.  Personally, I suspect that there are competent statisticians within Iran who could help them.  I know no names, but it would surprise me if everyone with any knowledge of statistics has fled the country.  Certainly I know a number of wonderful statisticians of Iranian origin who have done just that and with whom I have been honored to work with over the years, but is there no one left there? 

 

Obviously, these thoughts are mine own and have nothing to do with my employer.

 

--Dave Judkins

 

---

From: Survey Research Methods Section of the ASA [[hidden email]] on behalf of Amir Kasaeian [[hidden email]]
Sent: Monday, June 24, 2013 6:00 PM
To: [hidden email]
Subject: Re: Survey sampling weight calculation

I think I found what you describe. just to verify I check it again with you.

I have the following geographic entities:

 

1. Province (all provinces are sampled with the same sample size of 1000)

2. Cluster (in each province there are 50 clusters with the same size of 20 individuals in each of them (50*20=1000 individuals).)

3. District (all districts of each province are in the sample: at this stage the sample size is proportional to each district total population)

4. Area (Urban/Rural, at this stage it seems the sample size is also proportional to the total population of that area)

5. individuals.

 

so concerning your explanation my provinces are strata and clusters are clusters.

I understand from data that clusters are sampled in a way that cover all districts of each province but in each district the size of rural and urban areas are not proportionally. Instead, in some districts of a province all the individuals are from rural or urban areas only. So, I don't know the role of Districts and Areas in the weighing process??

concerning this information do you confirm the way of weighing you mentioned? does area type have any role in the weighting process or not?

 

 

My many many thanks your attention and favours guiding me. Please let me know if more information is needed.

 

Sincerely yours,

Amir

 

 

For better understanding I added some results below: 

. tab province

Province	Freq.	Percent	Cum.
1	1,000	3.33	3.33
2	1,000	3.33	6.66
3	1,000	3.33	9.99
4	995	3.31	13.30
5	1,001	3.33	16.64
6	1,000	3.33	19.97
7	1,002	3.34	23.30
8	1,020	3.40	26.70
9	999	3.33	30.03
10	1,001	3.33	33.36
11	1,000	3.33	36.69
12	1,000	3.33	40.02
13	1,000	3.33	43.35
14	1,000	3.33	46.68
15	1,000	3.33	50.01
16	1,000	3.33	53.34< /tr>
17	1,001	3.33	56.68
18	1,000	3.33	60.01
19	1,000	3.33	63.34
20	1,000	3.33	66.67
21	1,000	3.33	70.00
22	1,001	3.33	73.33
23	1,000	3.33	76.66
24	1,001	3.33	79.99
25	1,008	3.36	83.35
26	1,000	3.33	86.68
27	1,000	3.33	90.01
28	1,000	3.33	93.34
29	1,000	3.33	96.67
30	1,000	3.33	100.00
Total	30,029	100.00

no. of clusters of a given province:

. tab  cluster if  province_census==20

Cluster
Code	Freq.	Percent	Cum.
1	20	2.00	2.00
2	20	2.00	4.00
3	20	2.00	6.00
4	20	2.00	8.00
5	20	2.00	10.00
6	20	2.00	12.00
7	20	2.00	14.00
8	20	2.00	16.00
9	20	2.00	18.00
10	20	2.00	20.00
11	20	2.00	22.00
12	20	2.00	24.00
13	20	2.00	26.00
14	20	2.00	28.00
15	20	2.00	30.00
16	20	2.00	32.00
17	20	2.0034.00
18	20	2.00	36.00
19	20	2.00	38.00
20	20	2.00	40.00
21	20	2.00	42.00
22	20	2.00	44.00
23	20	2.00	46.00
24	20	2.00	48.00
25	20	2.00	50.00
26	20	2.00	52.00
27	20	2.00	54.00
28	20	2.00	56.00
29	20	2.00	58.00
30	20	2.00	60.00
31	20	2.00	62.00
32	20	2.00	64.00
33	20	2.00	66.00
34	20< /td>	2.00	68.00
35	20	2.00	70.00
36	20	2.00	72.00
37	20	2.00	74.00
38	20	2.00	76.00
39	20	2.00	78.00
40	20	2.00	80.00
41	20	2.00	82.00
42	20	2.00	84.00
43	20	2.00	86.00
44	20	2.00	88.00
45	20	2.00	90.00
46	20	2.00	92.00
47	20	2.00	94.00
48	20	2.00	96.00
49	20	2.00	98.00
50	20	2.00	100.00
Total	1,000	100.00

for a given province of 12 districts

.tab  district area if   province_census==23

District	Area Code
Code	Urban	Rural	Total
2	60	0	60
3	20	20	40
4	100	0	100
5	20	0	20
6	60	0	60
7	0	20	20
8	40	0	40
9	0	20	20
11	140	0	140
12	181	0	181
13	300	20	320
Total	921	80	1,001

for a optional province

. tab  idnum  x1d  if   province_census==20

 

	District	Code
Id Number	1	2	3	4	total
1	0	1	0	0	1
2	0	1	0	1	2
3	0	1	0	1	2
4	0	1	0	1	2
5	0	1	0	1	2
6	0	1	0	1	2
7	0	1	0	1	2
8	0	1	0	1	2
9	0	1	0	1	2
10	0	1	0	1	2
11	0	1	0	1	2
12	0	1	0	1	2
13	0	1	0	1	2
14	0	1	0	1	2
15	0	1	0	1	2
16	0	1	0	1	2
17	0	1	0	1	2
18	0	1	0	1	2
19	0	1	0	1	2
20	0	1	0	1	2
21	0	1	0	1	2
22	0	1	0	1	2
23	0	1	0	1	2
24	0	1	0	1	2
25	0	1	0	1	2
26	0	1	0	1	2
27	0	1	0	1	2
28	0	1	0	1	2
29	0	1	0	1	2
30	0	1	0	1	2
31	0	1	0	1	2
32	0	1	0	1	2
33	0	1	0	1	2
34	0	1	0	1	2
35	0	1	0	1	2
36	0	1	0	1	2
37	0	1	0	1	2
38	0	1	0	1	2
39	0	1	0	1	2
40	0	1	0	1	2
41	0	1	0	1	2
42	0	1	0	1	2
43	0	1	0	1	2
44	0	1	0	1	2
45	0	1	0	1	2
46	0	1	0	1	2
47	0	1	0	1	2
48	0	1	0	1	2
49	0	1	0	1	2
50	0	1	0	1	2
51	0	1	0	1	2
52	0	1	0	1	2
53	0	1	0	1	2
54	0	1	0	1	2
55	0	1	0	1	2
56	0	1	0	1	2
57	0	1	0	1	2
58	0	1	0	1	2
59	0	1	0	1	2
60	0	1	0	1	2
61	1	0	0	1	2
62	1	0	0	1	2
63	1	0	0	1	2
64	1	0	0	1	2
65	1	0	0	1	2
66	1	0	0	1	2
67	1	0	0	1	2
68	1	0	0	1	2
69	1	0	0	1	2
70	1	0	0	1	2
71	1	0	0	1	2
72	1	0	0	1	2
73	1	0	0	1	2
74	1	0	0	1	2
75	1	0	0	1	2
76	1	0	0	1	2
77	1	0	0	1	2
78	1	0	0	1	2
79	1	0	0	1	2
80	1	0	0	1	2
81	1	0	0	1	2
82	1	0	0	1	2
83	1	0	0	1	2
84	1	0	0	1	2
85	1	0	0	1	2
86	1	0	0	1	2
87	1	0	0	1	2
88	1	0	0	1	2
89	1	0	0	1	2
90	1	0	0	1	2
91	1	0	0	1	2
92	1	0	0	1	2
93	1	0	0	1	2
94	1	0	0	1	2
95	1	0	0	1	2
96	1	0	0	1	2
97	1	0	0	1	2
98	1	0	0	1	2
99	1	0	0	1	2
100	1	0	0	1	2
101	1	0	0	1	2
102	1	0	0	1	2
103	1	0	0	1	2
104	1	0	0	1	2
105	1	0	0	1	2
106	1	0	0	1	2
107	1	0	0	1	2
108	1	0	0	1	2
109	1	0	0	1	2
110	1	0	0	1	2
111	1	0	0	1	2
112	1	0	0	1	2
113	1	0	0	1	2
114	1	0	0	1	2
115	1	0	0	1	2
116	1	0	0	1	2
117	1	0	0	1	2
118	1	0	0	1	2
119	1	0	0	1	2
120	1	0	0	1	2
121	1	0	0	1	2
122	1	0	0	1	2
123	1	0	0	1	2
124	1	0	0	1	2
125	1	0	0	1	2
126	1	0	0	1	2
127	1	0	0	1	2
128	1	0	0	1	2
129	1	0	0	1	2
130	1	0	0	1	2
131	1	0	0	1	2
132	1	0	0	1	2
133	1	0	0	1	2
134	1	0	0	1	2
135	1	0	0	1	2
136	1	0	0	1	2
137	1	0	0	1	2
138	1	0	0	1	2
139	1	0	0	1	2
140	1	0	0	1	2
141	1	0	0	1	2
142	1	0	0	1	2
143	1	0	0	1	2
144	1	0	0	1	2
145	1	0	0	1	2
146	1	0	0	1	2
147	1	0	0	1	2
148	1	0	0	1	2
149	1	0	0	1	2
150	1	0	0	1	2
151	1	0	0	1	2
152	1	0	0	1	2
153	1	0	0	1	2
154	1	0	0	1	2
155	1	0	0	1	2
156	1	0	0	1	2
157	1	0	0	1	2
158	1	0	0	1	2
159	1	0	0	1	2
160	1	0	0	1	2
161	0	0	1	1	2
162	0	0	1	1	2
163	0	0	1	1	2
164	0	0	1	1	2
165	0	0	1	1	2
166	0	0	1	1	2
167	0	0	1	1	2
168	0	0	1	1	2
169	0	0	1	1	2
170	0	0	1	1	2
171	0	0	1	1	2
172	0	0	1	1	2
173	0	0	1	1	2
174	0	0	1	1	2
175	0	0	1	1	2
176	0	0	1	1	2
177	0	0	1	1	2
178	0	0	1	1	2
179	0	0	1	1	2
180	0	0	1	1	2
181	0	0	1	1	2
182	0	0	1	1	2
183	0	0	1	1	2
184	0	0	1	1	2
185	0	0	1	1	2
186	0	0	1	1	2
187	0	0	1	1	2
188	0	0	1	1	2
189	0	0	1	1	2
190	0	0	1	1	2
191	0	0	1	1	2
192	0	0	1	1	2
193	0	0	1	1	2
194	0	0	1	1	2
195	0	0	1	1	2
196	0	0	1	1	2
197	0	0	1	1	2
198	0	0	1	1	2
199	0	0	1	1	2
200	0	0	1	1	2
201	0	0	1	1	2
202	0	0	1	1	2
203	0	0	1	1	2
204	0	0	1	1	2
205	0	0	1	1	2
206	0	0	1	1	2
207	0	0	1	1	2
208	0	0	1	1	2
209	0	0	1	1	2
210	0	0	1	1	2
211	0	0	1	1	2
212	0	0	1	1	2
213	0	0	1	1	2
214	0	0	1	1	2
215	0	0	1	1	2
216	0	0	1	1	2
217	0	0	1	1	2
218	0	0	1	1	2
219	0	0	1	1	2
220	0	0	1	1	2
221	0	0	1	1	2
222	0	0	1	1	2
223	0	0	1	1	2
224	0	0	1	1	2
225	0	0	1	1	2
226	0	0	1	1	2
227	0	0	1	1	2
228	0	0	1	1	2
229	0	0	1	1	2
230	0	0	1	1	2
231	0	0	1	1	2
232	0	0	1	1	2
233	0	0	1	1	2
234	0	0	1	1	2
235	0	0	1	1	2
236	0	0	1	1	2
237	0	0	1	1	2
238	0	0	1	1	2
239	0	0	1	1	2
240	0	0	1	1	2
241	0	0	1	1	2
242	0	0	1	1	2
243	0	0	1	1	2
244	0	0	1	1	2
245	0	0	1	1	2
246	0	0	1	1	2
247	0	0	1	1	2
248	0	0	1	1	2
249	0	0	1	1	2
250	0	0	1	1	2
251	0	0	1	1	2
252	0	0	1	1	2
253	0	0	1	1	2
254	0	0	1	1	2
255	0	0	1	1	2
256	0	0	1	1	2
257	0	0	1	1	2
258	0	0	1	1	2
259	0	0	1	1	2
260	0	0	1	1	2
261	0	0	1	1	2
262	0	0	1	1	2
263	0	0	1	1	2
264	0	0	1	1	2
265	0	0	1	1	2
266	0	0	1	1	2
267	0	0	1	1	2
268	0	0	1	1	2
269	0	0	1	1	2
270	0	0	1	1	2
271	0	0	1	1	2
272	0	0	1	1	2
273	0	0	1	1	2
274	0	0	1	1	2
275	0	0	1	1	2
276	0	0	1	1	2
277	0	0	1	1	2
278	0	0	1	1	2
279	0	0	1	1	2
280	0	0	1	1	2
281	0	0	1	1	2
282	0	0	1	1	2
283	0	0	1	1	2
284	0	0	1	1	2
285	0	0	1	1	2
286	0	0	1	1	2
287	0	0	1	1	2
288	0	0	1	1	2
289	0	0	1	1	2
290	0	0	1	1	2
291	0	0	1	1	2
292	0	0	1	1	2
293	0	0	1	1	2
294	0	0	1	1	2
295	0	0	1	1	2
296	0	0	1	1	2
297	0	0	1	1	2
298	0	0	1	1	2
299	0	0	1	1	2
300	0	0	1	1	2
301	0	0	1	1	2
302	0	0	1	1	2
303	0	0	1	1	2
304	0	0	1	1	2
305	0	0	1	1	2
306	0	0	1	1	2
307	0	0	1	1	2
308	0	0	1	1	2
309	0	0	1	1	2
310	0	0	1	1	2
311	0	0	1	1	2
312	0	0	1	1	2
313	0	0	1	1	2
314	0	0	1	1	2
315	0	0	1	1	2
316	0	0	1	1	2
317	0	0	1	1	2
318	0	0	1	1	2
319	0	0	1	1	2
320	0	0	1	1	2
321	0	0	1	1	2
322	0	0	1	1	2
323	0	0	1	1	2
324	0	0	1	1	2
325	0	0	1	1	2
326	0	0	1	1	2
327	0	0	1	1	2
328	0	0	1	1	2
329	0	0	1	1	2
330	0	0	1	1	2
331	0	0	1	1	2
332	0	0	1	1	2
333	0	0	1	1	2
334	0	0	1	1	2
335	0	0	1	1	2
336	0	0	1	1	2
337	0	0	1	1	2
338	0	0	1	1	2
339	0	0	1	1	2
340	0	0	1	1	2
341	0	1	0	1	2
342	0	1	0	1	2
343	0	1	0	1	2
344	0	1	0	1	2
345	0	1	0	1	2
346	0	1	0	1	2
347	0	1	0	1	2
348	0	1	0	1	2
349	0	1	0	1	2
350	0	1	0	1	2
351	0	1	0	1	2
352	0	1	0	1	2
353	0	1	0	1	2
354	0	1	0	1	2
355	0	1	0	1	2
356	0	1	0	1	2
357	0	1	0	1	2
358	0	1	0	1	2
359	0	1	0	1	2
360	0	1	0	1	2
361	0	1	0	1	2
362	0	1	0	1	2
363	0	1	0	1	2
364	0	1	0	1	2
365	0	1	0	1	2
366	0	1	0	1	2
367	0	1	0	1	2
368	0	1	0	1	2
369	0	1	0	1	2
370	0	1	0	1	2
371	0	1	0	1	2
372	0	1	0	1	2
373	0	1	0	1	2
374	0	1	0	1	2
375	0	1	0	1	2
376	0	1	0	1	2
377	0	1	0	1	2
378	0	1	0	1	2
379	0	1	0	1	2
380	0	1	0	1	2
381	0	1	0	1	2
382	0	1	0	0	1
383	0	1	0	1	2
384	0	1	0	1	2
385	0	1	0	1	2
386	0	1	0	1	2
387	0	1	0	1	2
388	0	1	0	1	2
389	0	1	0	1	2
390	0	1	0	1	2
391	0	1	0	1	2
392	0	1	0	1	2
393	0	1	0	1	2
394	0	1	0	1	2
395	0	1	0	1	2
396	0	1	0	1	2
397	0	1	0	1	2
398	0	1	0	1	2
399	0	1	0	1	2
400	0	1	0	1	2
401	0	1	0	1	2
402	0	1	0	1	2
403	0	1	0	0	1
404	0	1	0	0	1
405	0	1	0	0	1
406	0	1	0	0	1
407	0	1	0	0	1
408	0	1	0	0	1
409	0	1	0	0	1
410	0	1	0	0	1
411	0	1	0	0	1
412	0	1	0	0	1
413	0	1	0	0	1
414	0	1	0	0	1
415	0	1	0	0	1
416	0	1	0	0	1
417	0	1	0	0	1
418	0	1	0	0	1
419	0	1	0	0	1
420	0	1	0	0	1
421	0	1	0	0	1
422	0	1	0	0	1
423	0	1	0	0	1
424	0	1	0	0	1
425	0	1	0	0	1
426	0	1	0	0	1
427	0	1	0	0	1
428	0	1	0	0	1
429	0	1	0	0	1
430	0	1	0	0	1
431	0	1	0	0	1
432	0	1	0	0	1
433	0	1	0	0	1
434	0	1	0	0	1
435	0	1	0	0	1
436	0	1	0	0	1
437	0	1	0	0	1
438	0	1	0	0	1
439	0	1	0	0	1
440	0	1	0	0	1
441	0	1	0	0	1
442	0	1	0	0	1
443	0	1	0	0	1
444	0	1	0	0	1
445	0	1	0	0	1
446	0	1	0	0	1
447	0	1	0	0	1
448	0	1	0	0	1
449	0	1	0	0	1
450	0	1	0	0	1
451	0	1	0	0	1
452	0	1	0	0	1
453	0	1	0	0	1
454	0	1	0	0	1
455	0	1	0	0	1
456	0	1	0	0	1
457	0	1	0	0	1
458	0	1	0	0	1
459	0	1	0	0	1
460	0	1	0	0	1
461	0	1	0	0	1
462	0	1	0	0	1
463	0	1	0	0	1
464	0	1	0	0	1
465	0	1	0	0	1
466	0	1	0	0	1
467	0	1	0	0	1
468	0	1	0	0	1
469	0	1	0	0	1
470	0	1	0	0	1
471	0	1	0	0	1
472	0	1	0	0	1
473	0	1	0	0	1
474	0	1	0	0	1
475	0	1	0	0	1
476	0	1	0	0	1
477	0	1	0	0	1
478	0	1	0	0	1
479	0	1	0	0	1
480	0	1	0	0	1
481	0	1	0	0	1
482	0	1	0	0	1
483	0	1	0	0	1
484	0	1	0	0	1
485	0	1	0	0	1
486	0	1	0	0	1
487	0	1	0	0	1
488	0	1	0	0	1
489	0	1	0	0	1
490	0	1	0	0	1
491	0	1	0	0	1
492	0	1	0	0	1
493	0	1	0	0	1
494	0	1	0	0	1
495	0	1	0	0	1
496	0	1	0	0	1
497	0	1	0	0	1
498	0	1	0	0	1
499	0	1	0	0	1
500	0	1	0	0	1
501	0	1	0	0	1
502	0	1	0	0	1
503	0	1	0	0	1
504	0	1	0	0	1
505	0	1	0	0	1
506	0	1	0	0	1
507	0	1	0	0	1
508	0	1	0	0	1
509	0	1	0	0	1
510	0	1	0	0	1
511	0	1	0	0	1
512	0	1	0	0	1
513	0	1	0	0	1
514	0	1	0	0	1
515	0	1	0	0	1
516	0	1	0	0	1
517	0	1	0	0	1
518	0	1	0	0	1
519	0	1	0	0	1
520	0	1	0	0	1
521	0	1	0	0	1
522	0	1	0	0	1
523	0	1	0	0	1
524	0	1	0	0	1
525	0	1	0	0	1
526	0	1	0	0	1
527	0	1	0	0	1
528	0	1	0	0	1
529	0	1	0	0	1
530	0	1	0	0	1
531	0	1	0	0	1
532	0	1	0	0	1
533	0	1	0	0	1
534	0	1	0	0	1
535	0	1	0	0	1
536	0	1	0	0	1
537	0	1	0	0	1
538	0	1	0	0	1
539	0	1	0	0	1
540	0	1	0	0	1
541	0	1	0	0	1
542	0	1	0	0	1
543	0	1	0	0	1
544	0	1	0	0	1
545	0	1	0	0	1
546	0	1	0	0	1
547	0	1	0	0	1
548	0	1	0	0	1
549	0	1	0	0	1
550	0	1	0	0	1
551	0	1	0	0	1
552	0	1	0	0	1
553	0	1	0	0	1
554	0	1	0	0	1
555	0	1	0	0	1
556	0	1	0	0	1
557	0	1	0	0	1
558	0	1	0	0	1
559	0	1	0	0	1
560	0	1	0	0	1
561	0	1	0	0	1
562	0	1	0	0	1
563	0	1	0	0	1
564	0	1	0	0	1
565	0	1	0	0	1
566	0	1	0	0	1
567	0	1	0	0	1
568	0	1	0	0	1
569	0	1	0	0	1
570	0	1	0	0	1
571	0	1	0	0	1
572	0	1	0	0	1
573	0	1	0	0	1
574	0	1	0	0	1
575	0	1	0	0	1
576	0	1	0	0	1
577	0	1	0	0	1
578	0	1	0	0	1
579	0	1	0	0	1
580	0	1	0	0	1
581	0	1	0	0	1
582	0	1	0	0	1
583	0	1	0	0	1
584	0	1	0	0	1
585	0	1	0	0	1
586	0	1	0	0	1
587	0	1	0	0	1
588	0	1	0	0	1
589	0	1	0	0	1
590	0	1	0	0	1
591	0	1	0	0	1
592	0	1	0	0	1
593	0	1	0	0	1
594	0	1	0	0	1
595	0	1	0	0	1
596	0	1	0	0	1
597	0	1	0	0	1
598	0	1	0	0	1
599	0	1	0	0	1
600	0	1	0	0	1
Total	100	320	180	400	1,000

 

Amir Kasaeian,
PhD Candidate in Biostatistics,

Dept. of Epidemiology and Biostatistics,

School of Public Health,

Tehran University of Medical Sciences (TUMS).

P.B. : 14155-6446

Cell Phone: [phone]
E-mail: [hidden email]

            [hidden email]  

 

---

From: Stas Kolenikov <[hidden email]>
To: [hidden email]
Sent: Tuesday, 25 June 2013, 0:10
Subject: Re: Survey sampling weight calculation

So you have the following geographic entities:
1. province;
2. district;
3. block (I would rather give it a different name, as "cluster" is a
sampling term);
4. individual

At each level, you need to determine whether it is this a stratum, a
cluster, or a domain. Stratum means it predetermined before the random
draws were done. If the country that you are studying has 30
provinces, and samples were taken in each province independently, then
these are strata. Cluster means there was a list of units, and some
were sampled from that list. If there are 75 provinces, and 30 were
drawn randomly, you need to know the procedure that was used to draw
them. There may have been 75 provinces, and 30 were selected by the
investigator for a specific reason, but they could have been pooled
together, in which case they form domains. To determine probabilities
of selection, you need to know how many units of this level were there
to sample from: how many provinces in the country; how many blocks in
the province or district (if the blocks were selected from districts);
how many individuals in each cluster.

If you don't have the detailed numbers (e.g., the field organization
discarded their records at the cluster level, so there is no cluster
information), then the best you can do is to take the probability of
selection to be 1000/population of the province, and use blocks as the
primary sampling units (if provinces act as strata):

bysort province (cluster individual) : gen approx_prob_weight =
province_population / _N
svyset cluster [pw=approx_prob_weight], strata( province )

This is just an example of how you could set this up; my guess is that
this is how your sample was designed, but, once again, you need to
verify what the procedure was with whoever did it.


-- Stas Kolenikov, PhD, PStat (SSC)
-- Senior Survey Statistician, Abt SRBI
-- Opinions stated in this email are mine only, and do not reflect the
position of my employer
-- http://stas.kolenikov.name



On Mon, Jun 24, 2013 at 12:38 PM, Amir Kasaeian <[hidden email]> wrote:

> Dear all,
> I have question about calculating survey weights for my data.
> Suppose the survey covers 30 provinces with different population and in each
> province 1000 individuals are sampled.  In each province there are 50
> clusters with 20 individuals in each (50*20=1000).
> Also, in each province there are 1 to 20 districts. And from the frequency
> tables, gender and age groups frequencies (5 yrs groups) are the same or
> approximately similar in each district. There are also urban and rural areas
> in each district with the same frequencies of gender and age groups in each.
> My question is how to weight this survey now. I will be pleased to let me
> know how to calculate the survey weights by Stata.
> Your comments and help will be appreciated.
> Thank you so much.
>
> Yours,
> Amir
>
> Amir Kasaeian,
> PhD Candidate in Biostatistics,
> Dept. of Epidemiology and Biostatistics,
> School of Public Health,
> Tehran University of Medical Sciences (TUMS).
> P.B. : 14155-6446
> Cell Phone: [phone]
> E-mail: [hidden email]
>            [hidden email]
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ To subscribe/unsubscribe SRMSNet:
> http://listserv.umd.edu/cgi-bin/wa?A0=srmsnet&D=0&F=&H=0&O=T&S=&T=1
>
> SRMS website: http://www.amstat.org/sections/srms/
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
To subscribe/unsubscribe SRMSNet:
  http://listserv.umd.edu/cgi-bin/wa?A0=srmsnet&D=0&F=&H=0&O=T&S=&T=1

SRMS website:  http://www.amstat.org/sections/srms/
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ To subscribe/unsubscribe SRMSNet: http://listserv.umd.edu/cgi-bin/wa?A0=srmsnet&D=0&F=&H=0&O=T&S=&T=1

SRMS website: http://www.amstat.org/sections/srms/ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

 

---

This message may contain privileged and confidential information intended solely for the addressee. Please do not read, disseminate or copy it unless you are the intended recipient. If this message has been received in error, we kindly ask that you notify the sender immediately by return email and delete all copies of the message from your system.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ To subscribe/unsubscribe SRMSNet: http://listserv.umd.edu/cgi-bin/wa?A0=srmsnet&D=0&F=&H=0&O=T&S=&T=1

SRMS website: http://www.amstat.org/sections/srms/ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~