Multiple sampling methods / Techniques in data collection in social science are

Pahlaj  -  

I think that the social as well as medical sciences seem to often be complicated by the many variables involved. Thus multiple regression may be used, but it may be easy to misconstrue results. If you can keep things simple, then you could look for specific influences, such as in the first link below, which was written for another purpose. But still, you have to consider that even in the simplest situation, there may be other influences at work. However, yes it would be good to apply statistics using various techniques, and a good deal of 'common sense'/logic in evaluating each question you may wish to explore.

However, I am not certain what is meant by "multiple sampling methods."  Some techniques cannot be mixed. You need to be sure that your methods are logical.  However, multiple studies, conducted differently, may be helpful, if affordable. 

One caution I have, which is against tradition, is that I think that hypothesis testing is overused and misused.  One needs to be aware that no point hypothesis is exactly correct, and since a p-value is a function of sample size, gathering enough data at a given level will reject any null hypothesis. That is the subject of the second link below. So I suggest using confidence intervals, instead of hypothesis testing, whenever possible.   -  Thanks - Jim   

Technical Report A Note on Estimating Impact of Data Collection Mode Change f...

Article Practical Interpretation of Hypothesis Tests - letter to the...

Dear Pahlaj, Dear All,

The link below is unique in pointing out when a method of sampling is best. I hope that it helps,  My regards.

http://changingminds.org/explanations/research/sampling/choosing_sampling.htm

Hi. To Nizar's attachment, I'd like to add model-assisted design-based sampling. (If it was there, I missed it, sorry.)  The modeling part helps if there happens to be a random selection, that by bad luck, is not very 'representative.'  Therefore it combines the features of randomization and prediction. Also to Nizar's reference, I'd like to add these: 

Sampling book references:

Cochran, W.G.(1977), Sampling Techniques, 3rd ed., John Wiley & Sons.

Särndal, C.-E., Swensson, B. and Wretman, J. (1992), Model Assisted Survey Sampling, Springer-Verlag.

Brewer, KRW (2002), Combined survey sampling inference: Weighing Basu's elephants, Arnold: London and Oxford University Press.

Lohr, S.L.(2010), Sampling: Design and Analysis, 2nd ed., Brooks/Cole.

Cochran and Lohr are perhaps the most 'user friendly.' 

I did not include a book on prediction, because the only one of which I am aware is not at all 'user friendly.'   There are a number of papers on it.  I did a lot of them covering cutoff sampling.  I'll link something here.   - Thanks - Jim   

Article Efficacy of Quasi-Cutoff Sampling and Model-Based Estimation...

Also, note that sampling and estimation are related, such that your sampling method will dictate what estimation choices you have, unless you estimate using prediction. Also please note that stratification can be extremely important.       

Multiple sampling methods / Techniques in data collection in social science are

It is known that there are three types of sampling; single sampling, double sampling, and multiple sampling. Multiple sampling is an extension of double sampling, which involves the inspection of 1 to k successive samples.

There is also what is called, the multi-stage sample, which is one in which sampling is done sequentially across two or more hierarchical levels.

Yes, it is appropriate dear @Pahlaj to use multiple sampling methods in social science! How about Mixed Method!

Mixed Methods Sampling - A Typology With Example is a fine paper for this thread!

http://www.sagepub.com/bjohnsonstudy/articles/Teddlie.pdf

The link above, provided by Ljubomir certainly sounds like it answers the question, and got my up vote. There is, however, one area that concerns me. On page 90 of that paper, they mention "purposive random sampling," which I did not understand. I only skimmed this, but I did not see how the data are to be evaluated, or exactly what this was.  That is, inference can be made from random samples, and from purposive IF there are regressor data on the population, but estimations for a 'purposive random sample' would need to be predictions using regressor data and only randomizing to help, as in stratification, with bias reduction.  (Stratification also can help with variance reduction.)  Perhaps I don't understand, and as in quota sampling, no general inference is made, just some subjective, descriptive information.  Thanks.  

James:  I only scanned the doc you are referring to as well.  However, I think it is also important to note that the Purposive Sampling is rarely used in the generation of Quantitative data according to Table 3 on page 87 and more often used in qualitative analysis where sampling is determined by a lead investigator, by definition, is permitted to inject his/her subject matter expertise into the selection of population representative sample study points. The introduction of potential bias seems apparent but acceptable if basic conventions of analysis are followed. These may be implied in the actual study but not spelled out in the paper.  I am facing a small sample size on my project in Pakistan (about 30) when I had hoped for more than 200 in order to perform SEM.  But the general unrest at the borders and the time constraints on the national representatives working with me on the project there have all but forced me into looking for new sampling/analysis alternatives. 'Guestimating' from the sample size of 12 in the example on p. 90, the value in the small sample size is the long-term observation of selected sites or target sites which could then build over time and potentially produce more statistical quantitative analysis data.  Certainly the intermittent qualitative analysis could be most informative and act like a foundation similar to qualitative techniques such as Phenominological Research studies where case studies can evolve in multiple directions but may converge on certain points of interest over a number of samples at multiple periods of time. Great information in this thread gentlemen...additional comments welcome.

Dear All,

Thank you for spending your valuable time in reading my question, and searching for resources to share with us, as well as offering useful opinions and insights. 

Now, we are confident that we can use combinations (mixed) of methods to get samples from populations, our population includes groups, sub-groups, and individuals, as well as key informants. We now plan to use stratified random sampling to select samples from each strata (various group, subgroup or individuals) to find answer whether the intervention has any significant effect, as well as we plan to use purposive sampling to get samples from key informants to get data on commodity value chain.

We plan to use quantitative as well as qualitative methods to get answer of our questions on impact assessment - whether joining groups has any significant impact of intervention, etc. 

We still hope to hear of your expert opinions and receive your valuable suggestions on our approach. 

Perhaps I missed scale of sampling, e.g. geographic coordinates of the study subjects? If the questions involved humans in general, sampling should be conducted throughout the world, or not?

Provided your research question(s) is answerable with more than one method, the more the method(s), the better your results, especially if the methods are combined well. Social sciences methods/researchers are becoming more flexible than they used to be so carefully see if you can combine methods in this study. As someone suggested, many articles in Journal of Mixed Method Research will help, but you can also see John Creswell's book: Research design: Qualitative, Quantitative and Mixed methods approaches.  

Acceptance sampling plans determine the sample size and criteria for accepting or rejecting whole the batch. Organizations having ISO certificate should follow the procedures of the acceptance sampling.

"Acceptance sampling involves both the producer (or supplier) of materials and the consumer (or buyer). Consumers need acceptance sampling to limit the risk of rejecting good-quality materials or accepting bad-quality materials. Consequently, the consumer, sometimes in conjunction with the producer through contractual agreements, specifies the parameters of the plan. Any company can be both a producer of goods purchased by another company and a consumer of goods or raw materials supplied by another company."

More information may be obtained through the following link:

http://wps.prenhall.com/wps/media/objects/7117/7288732/65767_28_SuppG.pdf

The above answers are all satisfactory. However, they failed to take note of "purposive random sampling" described in http://sociologyofeurope.unifi.it/upload/sub/documenti/Teddlie%20-%20Mixed%20Methods%20Sampling%20-%20A%20Typology%20With%20Examples.pdf

Journal of Mixed Methods Research

DOI: 10.1177/2345678906292430

Journal of Mixed Methods Research 2007; 1; 77

Charles Teddlie and Fen Yu

Mixed Methods Sampling: A Typology With Examples

http://mmr.sagepub.com/cgi/content/abstract/1/1/77

The online version of this article can be found at:

Published by:

http://www.sagepublications.com

Additional services and information for Journal of Mixed Methods Research can be found at:

Email Alerts: http://mmr.sagepub.com/cgi/alerts

Subscriptions: http://mmr.sagepub.com/subscriptions

Reprints: http://www.sagepub.com/journalsReprints.nav

Permissions: http://www.sagepub.com/journalsPermissions.nav