Before considering sample size, it is essential to think about the nature of your research, your goals, resources and the social context where you are working.

Are you thinking in a exploratory research, or, perhaps are you trying to develop an explanatory study.

Are you considering a qualitative or quantitative approach (or both)?.

Without such kind of questions, it is hard to say: "You must apply this formula to determinate the sample size"

Regards

Leopoldo Medina

Use a sample size calculator,  there are many of these on line, just type in "sample size calculator"   in your search and then select one ie http://www.raosoft.com/samplesize.html ,  there will margin of error, confidence interval boxes, fill in the desired %  and  population size it will calculate this information for you.  

Concerning sample size, it boils down to the nature of your research questions.

Mason, M (2010). Sample Size and Saturation in PhD Studies Using Qualitative Interviews [63 paragraphs]. Forum Qualitative Sozialforschung / Forum: Qualitative Social Research, 11(3): 1- 8.

Sample size in Social Science Research also depending on what Research Method you are adopting e.g. Qualitative Research, Quantitative Research, Covariance-Based Structural Equation Modeling (SEM), Partial Least Squares SEM etc or even you are merely doing a pilot study / pre-test.

You might want to refer to this RG link that many scholars had shared their view points pertaining to sample size.  Wishing you all the best.

https://www.researchgate.net/topic/sample_size

There is no lower limit. Although if you only use one persons account I would hardly call this a sample. The size depends on what you want to do with the data. Qualitative projects will often look to have about 20 interviews, although an indepth analysis could use only a few. The simplest quantitative analysis (descriptive) you need at least thirty cases I would say. Reply with more detail and I'll try and help    

Aim of research in social sciences is to generalize the results  of sample population to target or conceptual population. so sample should be true representation of population under study.

Now coming to the question, for size of sample depends upon the population size( most important), variety, and geographical position and access are the  points to be considered for sample size.

if you are studying leadership styles of prime ministers or presidents  of a country then you will select all because of lesser in number.

if you are interested to study the  behavior of political leaders then sample size will be more.

if you go for consumers of some famous brand then you will incorporate variety and size taken from company.

if you study the living conditions of people living in deserts you need to access depending upon your courage, accessibility  to scattered population.

Lastly, type of your study will give you fair idea about sample size.

Good Luck

Rafique

Hi Rafique, I'm not sure where you got that from. I assure you the object of social science is NOT to generalize the results of a sample! 

Hi Paul,

I think it is like this but you are also right  as it is not like this  always.  Somehow every  study is  an input to decision makers and it depends upon what is the position of decision makers about which they want to take decisions.

It is preferable that the  study be generalized to at least one set of population.

 Please have a look at a book on research methods by  Umma Sekaran, page no. 25 ( 4th edition)

Regards 

I would echo Brien Bolin in searching for an online tool to assist in calculating your sample size. These are great for large-scale questionnaires or for calculating a representative sample of large population. 

In addition I would also reiterate David Fields assertion that sample size depends on the type of research you are conducting and the methodology(ies) you are employing. For quantitative research, sample size can be easily determined through a calculator as mentioned above. However for qualitative research the approach may be a little different. If you are conducting focus groups you may not be able to determine the appropriate sample size until you start noticing repetition of answers among your focus group, or saturation. Therefore, it all depends. As for specific reference to text, any sociological research methods textbook will have a section on sampling.

Hope this was somewhat helpful

Cheers

Dear Muhammad:

You must understand that in social sciences there are two kind of inferences

1. Theoretical - With your data You can postulate new causal relationships between variables. In this case, you don't need a big N sample. For example, take the Ph.D dissertation of Mark Granovetter, with a litte N, He has built a powerfull hypothesis in sociology: the strong of weak ties. This is the creative branch of science.

2. Probabilistic. This is the case when You are finding a generalization of results, a classic problem in economics or in demography. This is verificacionist branch of science.

All the best

Silvio Salej H.

Before considering sample size, it is essential to think about the nature of your research, your goals, resources and the social context where you are working.

Are you thinking in a exploratory research, or, perhaps are you trying to develop an explanatory study.

Are you considering a qualitative or quantitative approach (or both)?.

Without such kind of questions, it is hard to say: "You must apply this formula to determinate the sample size"

Regards

Leopoldo Medina

Leopoldo Medina's response above is astute. 

'Social science' is a vast category with many related methods (and methodologies). I'm sure people will help out if you could provide more detail of what you want to do.     

 Hi Silvio,

thanks for clarification and making me wise on the topic.

Regards

Muhammad Rafique

Agree with Leopoldo  Medina andPaul Whybrow.

Leopoldo's answer says it all. To give an example, if you want to research on the Jarawa tribes in Andaman Islands, their total population is between 250 to 400 and they speak one dialect. You may have to take the entire lot as the sample. But if you want to research on India, we are 1.2 billion; speak 150 languages and 1500 dialects and we are as diverse as you can imagine from one another every 50 odd miles. There fore, it is inaccurate to surmise that a sample of X number of people from a region will give credence to generalizations on Indians.

Muhammad:

I think you have received a lot of very relevant answers.

I think it could also be mentioned your comment about primary data.

What kind of primary data are you considering, what kind of variables are related to such data.

I think, for instance, in certain data, such as the answers about vote intention, in an electoral process. In this kind of situations, the sample size can be small and even so, forecasting can be very precise.

By other side, there are important discussion about the difference between statistical and sociocultural significance, it is to say that the difference or similitude recognized by statistics could be not coincide with nature and impact of several social phenomena.

Regards

Leopoldo Medina

its dependent to the topic and the magnitude of the academic degree one is going to get

Hi Faria Mirza,

I'm curious; what do you mean by 'the magnitude of the academic degree'? Why does have anything to do with the appropriate sample size?

Paul

oh i mean to say that is that msc, mphil. or phd for which we wana study? or a class assignment etc?

Hi Faria Mirza

While I see why you might say that, I have to disagree. Primarily because it implies that the larger the sample, the better or more prestigious the study. This is really not true and it is important not to make that mistake.

Some of the world-leading social scientists draws from a small sample. I just so happen to be reviewing a qualitative study that has been widely criticised for use too large a sample!  My PhD used a sample of >36,000, but it is not as valuable as my recent publication that draws on a sample of 21.

I think some of the difficulty here is that a lot of people think of sampling in the very narrow sense of seeking statistical representativeness. But there are many are other types of sampling.  The question of sample size is absolutely dependent on what you want to do with the data; regardless your academic level.

Unfortunately we don't know anything about  Israr Muhammad's study to give any more detail.

Paul

Based on the study objectives the sample size will be different. The alpha level, power of study, and cohen's d, one or two tail hypothesis at least needed to estimate the sample size.   If someone interested in finding a correlation or association between variables, for example,  10-15 subject per each independent/predictor  is enough as a rule of thumb.

It is usually 10% of the population size irrespective of the nature of the research as long as it not secondary data.

The very nature of sampling has become somehow irrelevant in spearhead science. In terms of data science, sampling corresponds to a world -and situations- in which little information was available. Nowadays, there is so much information at hand, that we cannot and we should not discard any kind.

Most cordially, I would invite you to spet forwards on to working with big data. Technically, this requieres either some ommand of R or of Python, two different suitable programming languages to work with big data.

The very nature and status of social sciences has changed.

And yet some things never change! There has been a lot of prophesying about the coming of an almighty 'big data' social science. This is the latest in a long history of revolutionary predictions with the school of thought known as social science. However what I think has been particularly interesting about this, from a sociological point of view, is the overt and unashamed connection of social scientific research with commercial interests. So whilst I agree that social scientists should pay attention to big data, it is rather because I think it demonstrates that now, more than ever before, a critical social science needed.  

I do know about this inflation concerning data science. Culturally speaking, the core of my argument is just about th need to incorporate computation as a tool in social science. Networks is one way. Big-data is another one. To name but a few.

Thanks to all Dears for helping and guiding on the issue

Wow, such a variety of answers! The one I found most interesting was the one that claimed that if I were doing research in China and wanted to generalize to the country's population, I would need about 130 million people in my sample! (10%). That's the high end of the range (assuming we didn't want to generalize to the entire world's population). The low was 1. May I conclude that we are at least 95% certain that the true answer is somewhere between 1 and 130 million? George Gallup would be rolling in his grave!  You may want to read: Cohen, J. (1992). A power primer. Psychological Bulletin, 112, 155-159. 

Smack to the point dear Jeff, smack to the point, indeed!

It will depend on the word count for your piece of work how much primary research you need to do. When I done my dissertation I used a focus group for qualitative research of 4 to 6 people and 2 qualitative interviews. However, the piece of work I was doing was only 10, 000 words which is quite small. Others in my year done small sample surveys of around 10 to 20 people for the data they required for their work. Again it depends on how big your piece of work is and what the objective or your article is. I have one fellow in my year and he sent out over 400 + questionnaires for his 10,000 word piece. It depends on what your aims and objectives are and what you are trying to prove or disprove in your article.

 I agree, it's depend on the size sample and population

Depends on the population and the magnitude of group differences you are trying to detect (do a power analysis)

Hi. How can i explain results that have been statistically significant but not reach the magnitude pre established at the sample size calculation? Variability?

For example i have found 17 ml of difference between two groups but the magnitude established at the sample size calculation were 50 ml. I have always believe that the results must be statistically significant just in case that the difference have been reached (50ml)

43

I'm of the opinion that sample size is a function of population, area of coverage, rationale behind the research and overall objective of the study

I find this discussion very interesting because it shows how subjective are some of the academic decisions our institutions make. My application for promotion was rejected on the basis that my sample size was small. I found this crazy but I could not control the situation. Surprisingly, the same articles were published in internationally recognised refereed journals. Thanks for this information. I have saved it for future use.

Dear Muhammand

I'm speaking as sociologist.

Firstly, it is not exact or proper to talk on sample when the fact is a network or graph. This is a hard issue that the statistical research has not elucidated well. O this issue, I know an article of Mark Grannovetter written in joint venture with a Sweden statistician (American Journal of Sociology), you can find it easily.  The problem is that the sample criteria, worked in that paper, is valid only in the dyadic level.

Secondly,If we take for granted the sample problem, Wasserman and Faust (1994) consider a matrix with 10 nodes as a mininum level for making a good structural analysis. This is valid, obviously, for primary data.

I hope that this short ideas could be useful for you.

All discussions, suggestions and arguments are equally valid for those who already at a stage of stable academic stage. For beginners (upto PhD) this discussion does not give crude answer. Of course nothing is certain in the world.

Paul Whybrow: being late to the party, I've just seen your "43" answer to the basic question that started this thread. Well done! (I wonder how many readers made the connection.) :)

Paul Whybrow: being late to the party, I've just seen your "43" answer to the basic question that started this thread. Well done! (I wonder how many readers made the connection.)

Paul Whybrow, being late to the party, I've just seen your "43" answer to the basic question that started this thread. Well done! (I wonder how many readers made the connection.)

Paul Whybrow, being late to the party, I've just seen your "43" answer to the basic question that started this thread. Well done! I wonder how many readers made the connection.

Paul Whybrow, being late to the party, I've just seen your 43 answer to the basic question that started this thread. Well done! I wonder how many readers made the connection.

I'm unsure if 43 is the right answer. I think that the correct answer is N+1, so I think that now there are enough cases in the sample.

I still like my answer best that it's somewhere between 1 and 130 million !

Thanks, Jeff - I am unsure why my smart-aleck note to Paul suddenly replicated itself 4 more times!

30 sample is appropriate for research.