I appreciate the request to comment, Rmm, but I don't think I know enough about your particular problem domain.

That's the thing about applying weights to survey respondents - making responses from one person, or a group of people, more important than those of another. You would do this if you have a legitimate reason to think that one group is severely under-represented in your sampling frame, or in your final sample. Or if you have a theoretical reason for giving greater value to the responses of some, and lesser value to others.

You need to have a theory, and/or good evidence, to support the use of weights in the first place and some ideas about how much those weights should apply.

Thinking about it some more, the purpose of your research is likely to be important too. If you're interested only in the value of real estate affected by floods then your weights may apply to the value associated with the people/organisations you survey. If you are interested in the effects on people's homes then you may minimise commercial real estate and apply weights based on the sizes of families.

Hello

I think you considered all the constraints. for better results collect sufficient data. The main constraint is the stakeholder and you consider all the stakeholders. All the best

Many people dont have adequate training, experience, expertise to identify, design, manage and make decisions concerning floods. Roads and bridges are damaged, dams fail, people are swept away, homes and businesses flood, efforts to modify flooding sometimes fail. There is a group of specialists that may know more than others, but not all civil engineers, hydrologists, geomorphologists, meteorologists, and geologists are informed specifically on storm and flood computation, design, risk, frequency, etc. Having a college degree or advanced education may or may not provide the technical training that is needed. And some individuals just develop an interest in flooding, but many don’t. Farmers of bottomlands, public safety officers, emergency rescue teams, and city managers are more apt to develop the interest. I think I would ask more pointed questions, such as 1. Would you feel comfortable in evaluating flood frequency and risk associated with urban design and management? 2. You want to get home but the road is flooded and water is moving, do you drive on fast, slow, or stop and wait, or consider if there are other routes that are probably not flooded? 3. Would you feel comfortable designing or managing floods for your community, or would you want a position with these duties, risk, liabilities? 4. Is information about floods important to your life and considered in decisions about safety, building a home, storm evacuation route, etc.

I think you might even find that our flood specialists and managers are challenged, and not overconfident due to the uncertainty and public safety responsibilities. We need to admit, facilities sometimes fail, evacuation and other warning not always heeded, property is damaged, crops lost, and people are injured or die. It’s a heavy load to carry. Can you rate how much of a stake or interest each respondent has - maybe let them rate themselves by their answers to direct questions.

Dear Hume F. Winzar sir,

thank you very much for your informative comment.

1. Giving weight to an individual is my personal feeling. That was developed when I saw my respondents. Fist of all, I do have a very limited number of people, but I predict a couple of specific persons in each group will give me a more correct picture of my scenario than others. Then I feel, that idea may dissolve due to the majority's idea. So give them a significance I think it is best to have a weighting schema for respondents. But I do not have any intergroup comparisons, so one group is never better than others. In such a situation do you agree with my argument to have a weight?

2. Really, I am searching for a theory but could not yet. Hence I need to find an acceptable weight schema based on a few attributes which may specific to each group.

For example, when I want to find the decision-makers' idea on a particular scenario, they will give me a better answer if they had come across same kind of situation earlier. Then persons with the same kind of experience my give higher weights. As well if a person with lots of experience but without the experience of same situation, may give a more accurate and acceptable idea than a person with less experience. So I have two attributes, experience in same kind of situation and service experience.

But I think I got your point, that is, the attributes should be defined based on the interest. So I'm rethinking what is my interest and what should I clear from each group.

I hope you will give your idea to my understanding.

thank you very much again.

Dear Mohammad Nasar sir,

thank you very much for your answer.

Yes I will try my best to collect data from all possible stakeholders

thank you again

Dear William F. Hansen , Sir,

Thank you very much for your comprehensive comment.

Yes, people's educational background may not reflects the expertise in the filed. The interest is the key to discover knowledge. You have given me a couple of sample questions to discover it, and based on such questions I could weight the people. - Kindly let me did I clearly understand it or not?

Meantime let me appreciate the way you point your idea

How to find the individual weights?

Use Calibration Questions to assess the quality of the estimators. Use the accuracy and precision of the answers to these calibration questions for calculating the weights.

Roger Cooke established a method for this: Structured Expert Judgment, also called Cookes Method or Delft Method. It is already used in hundreds of surveys and proven to give the best result compared to other forms of expert elicitation (especially in environmental studies). There is an excellent free edx course about this: Decision Making Under Uncertainty: Introduction to Structured Expert Judgment [1]

[1] https://courses.edx.org/courses/course-v1:DelftX+SEJ1x+1T2020/course/

Dear Michael Krisper Sir,

Thank you very much for the resource you provided. I'm going through it, and it is interesting

I do not have enough of the context on your study. Still, I would like to make a few observations, which probably you may know already. I think, we need to be certain why we are using the weights. In probability samples, weighting is necessitated when our primary interest is in interval estimation. This is where we don't want to go wrong. Sometimes, non-response, deliberate oversampling of some categories of respondents and inadequate sampling frame may result in unrepresentative samples. Weighting helps us adjust for non- response and overaampling etc so that the sample correctly reflects the population on key variables.

Survey in Social Research, 2014,(6th edition) provides a method of weighting on

single or multiple variables. If your prime interest is significance testing you might not need to weight the sample. My apologies for a delayed response, Pardeep.

R M M Pradeep -

Survey weights are the inverse of the probability of selection. The idea is that if you have a quantitative value, and you want to estimate, say for example, a mean or total, you can let the data you collect represent groups of data altogether comprising the total population. Is that what you mean by weights? I don't think it is. But I could be wrong.

Regression weights basically reduce the impact of data points on an estimated regression line based on an increase in sigma for the estimated residuals (heteroscedasticity), generally for larger predictions. I don't think you mean that either. But I could be wrong.

Calibration weights are modified survey weights. Modification is based on auxiliary data, potentially including regression weights. Don't think that's it. But could be wrong.

I've noticed that I think some people (erroneously I believe) refer to regression coefficients in multiple linear regression, as "weights." I doubt that any textbook would say this, but I can see why it might be thought. These are used as factors to then add together terms for prediction. (For a random variable, y, we "predict" which is just the word used instead of "estimate.") But regression coefficients, though used this way for prediction, do not show comparative weights for pieces of a total because the predictors (regresssors or independent variables) are in different units. (Some "standardize" but I see little use in this. Also, there are influences between "independent" variables which mean these pieces impact each other so that individual contributions are not that straightforward.) Anyway, I thought for a moment that this is what you meant. However, you said "...weight the participants..." so I guess not.

Perhaps you just mean, Who is most important as a participant to your survey? That would indicate you either want to give more credence to some individuals, or it looks like you might want to conduct four different surveys. Your conceptualization here is the likely problem. This is not clear.

I think that you need to more clearly state the data items you wish to collect, what you plan to do with these data, and especially clearly state your goals. You may find that this is not exactly understood to you yet, and benefit by considering in detail what it is that you plan to do.

Best wishes - Jim

PS - Reading more of your comments above, perhaps you do want to know "Who is most important as a participant to your survey?" But what you want to do about it is the question. If you notice my other comments are with regard to what one does with the information. And what is it that you want to learn, and how do you express this? - Cheers.

Michael Krisper notes a "Structured Expert Judgment" technique I've not heard about. I did run across Delphi back, I think, in the late 1970s, but my work since never encountered anything but the measurable. (It sounds like they used a form of "calibration." Does that mean using auxiliary data here too?) Michael's suggestion does sound like it might fit your needs. If you think about exactly what you want to do, and look into Michael's suggestion, maybe you can get to where you want to go.

I think before you make a decision on what information from which group must you collect, you need to state your RQ(s) and hypotheses. What information to include and what not to include gas to do with your research design. I mean it is a design issue and your design depends on your research questions.