How to weight the analysis by the actual population?

[Rabab Al hasni]

Dear Statalist,

I hope you are well.

In fact, I have used the method of random sampling to collect my data. Is it necessary to apply the survey data analysis with my data (categorical variables)?

At the beginning of the analysis stage, I have tried but I found it very complicated. In addition, because of the limited time that I have for completing my analysis, I did not use it.

Therefore, I would like to ask please how to weight the analysis by the actual population? Does this mean that I should apply the survey data analysis technique?

Below is a small example of the dataset:

* Example generated by -dataex-. For more info, type help dataex
clear
input int firm byte(PO_GEN Owner_EDU) float(Ent_size PO_EDU2) long Need_Funding
 1 0 1 1 0 1
 2 0 4 1 2 1
 3 0 3 3 1 1
 4 0 2 1 1 1
 5 0 4 1 2 1
 6 0 4 1 2 1
 7 1 2 3 1 1
 8 1 4 1 2 1
 9 0 4 4 2 1
10 0 4 3 2 1
11 0 3 3 1 1
12 0 4 1 2 1
13 0 4 3 2 1
14 1 4 1 2 1
15 0 3 4 1 1
16 0 2 3 1 1
17 1 4 4 2 1
18 0 3 1 1 1
19 0 4 3 2 1
20 0 4 4 2 1
21 0 4 4 2 1
22 0 2 4 1 1
23 0 2 4 1 1
24 0 4 1 2 1
25 0 4 3 2 1
26 1 4 3 2 1
27 0 4 3 2 1
28 0 4 3 2 1
29 1 4 3 2 1
30 0 3 3 1 1
31 0 3 4 1 1
32 0 2 4 1 1
33 0 4 3 2 1
34 0 4 3 2 1
35 0 3 3 1 1
36 1 1 1 0 1
37 1 1 1 0 1
38 1 4 3 2 1
39 1 3 1 1 1
40 1 1 1 0 1
41 1 4 1 2 1
42 0 3 1 1 1
43 1 2 1 1 1
44 1 2 1 1 1
45 0 2 1 1 1
46 1 2 1 1 1
47 1 4 3 2 1
48 0 4 3 2 1
49 0 2 1 1 1
50 1 4 1 2 1
51 0 3 1 1 0
52 0 2 3 1 0
53 1 4 1 2 1
54 0 3 1 1 1
55 1 2 1 1 1
56 1 1 1 0 1
57 1 2 1 1 1
58 1 3 1 1 1
59 0 1 1 0 0
60 1 2 1 1 0
end
label values PO_GEN gender
label def gender 0 "Male", modify
label def gender 1 "Female", modify
label values Owner_EDU form_qualification
label def form_qualification 1 "Do not have academic qualifications", modify
label def form_qualification 2 "General diploma degree or lower", modify
label def form_qualification 3 "Diploma degree", modify
label def form_qualification 4 "Bachelor degree", modify
label values Ent_size Ent_size
label def Ent_size 1 "Micro", modify
label def Ent_size 3 "small", modify
label def Ent_size 4 "Medium", modify
label values PO_EDU2 BO_EDU2
label def BO_EDU2 0 "not have academic qualification", modify
label def BO_EDU2 1 "lower level of formal qualification", modify
label def BO_EDU2 2 "higher level of academic qualification", modify
label values Need_Funding Need_Funding
label def Need_Funding 0 "Applied", modify
label def Need_Funding 1 "Not Applied", modify

Greatly appreciate your help

Best regards,
Rabab

[Richard Williams]

If, in fact, you just did simple random sampling, there is no need to use svy.

if, on the other hand, cases had differing probabilities of selection, e.g. one case had a 1 in 100 probability of being selected, while another had 1 in 1000 — then you would need to use pweights.

Things are further complicated if you used stratification or clustering when drawing your sample.

From what you say, I’m guessing you didn’t do any of that. If you had you would probably already know the answer to your question.

But perhaps I misjudge the situation. If you describe exactly how you drew your sample we might be able to advise you better.

[Rabab Al hasni]

Dear Richard Williams,

Many thanks for replying.

In fact, at first, I have defined my research population. Then I contact institutions that have relations with the research population to ensure the size of the real population. Once I got lists of firms with their contact number. I calculated the sample size that I should consider for my research (I think the calculation formula I applied is for Richard Giger). Then, I started to select the company randomly and through the phone I collect data or by delivering the questionnaire to their place.

The stratified and cluster sampling cannot be worked with my type and nature of research population. Therefore, I used simple random sampling.


Best regards,
Rabab