You can use G-power software available free of cost. You can also look at the attached paper for a summary of various methods (including popular rules of thumb) for calculating sample size.

http://www.gpower.hhu.de/en.html

Dear Hemal,

I am attaching the summary table related to various sample sizes. You can mail me for complete details as to how these have been arrived at:

Method for determining/deciding sample size                             Highest Sample Size                                                                                                                   recommended/calculated

1 Based on sample size determining tables                                                         384

2 Based on types of scales used in the instrument designing                          384

3 A-Priori Power Analysis Using G*Power                                                             400

4 Based on number of factors to be analyzed                           500 by Comrey and Lee

                                                                                                                450 as per rule of 10

Dear Hemal,

I suggest you my books (in Spanish):

Ruiz Espejo, Mariano (2013). Exactness of Inference in Finite Populations. Bubok. Madrid.

Ruiz Espejo, Mariano (2014). Foundations of Objective Statistical Inference. Lulu Press. Raleigh, NC.

http://www.bubok.es/libros/229317/EXACTITUD-DE-LA-INFERENCIA-EN-POBLACIONES-FINITAS

http://www.lulu.com/shop/mariano-ruiz-espejo/fundamentos-de-la-inferencia-estad%C3%ADstica-objetiva/paperback/product-21821572.html

You can use G-power software available free of cost. You can also look at the attached paper for a summary of various methods (including popular rules of thumb) for calculating sample size.

http://www.gpower.hhu.de/en.html

Hemal -

It depends upon the kind of data you are considering.  Would your research be in the area of continuous data?  Would it be yes/no data for proportions?  Are you essentially looking at infinite populations or finite populations?   There is a chapter on sample size estimation in each of these two books, and many others, I'd imagine:

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

Blair, E. and Blair, J(2015), Applied Survey Sampling, Sage Publications.

Also, I don't own the following book to confirm that it has a chapter on sample size requirements, but I have used it before at work, and it is excellent in that it covers a great deal of information in a very easy, well-written style:

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

You mention "sample size calculation," and that makes me wonder if you are talking about the "calculators" that you may find on the internet.  If so, please note that those you find might be assuming yes/no data for an infinite population, considering a worst case for sampling needs, i.e. when p=q=0.5. 

First, if you are looking at a finite population, it may be appropriate to consider a finite population correction (fpc) factor, which is used for both continuous data and for yes/no data.  (I am not familiar with other types, but I imagine this has some meaning for other data as well.) 

 https://www.researchgate.net/publication/262970761_FINITE_POPULATION_CORRECTION_%28fpc%29_FACTOR

This is because the focus is on variance estimation due to sampling errors, which means the error incurred because you did not have a census, and it just focuses on variance.  Indirectly it may include some bias and nonsampling error influence, but that is not what a standard error used in the estimation of sample size requirements is designed to do.  However, it can be very useful, depending upon application. 

Second, a sample size for yes/no results may assume a worst case for p and q, but if you have a preliminary estimate, you may find you need a substantially smaller sample size.  More is not always better if it strains your resources and perhaps adds to nonsampling error.  Application is important, as emphasis will vary.

What you really need for both continuous and yes/no data is a preliminary estimate of population standard deviation (standard error for estimated residuals in regression) which is an otherwise unknown constant for a given population, to be able to estimate a standard error for a mean or other parameter of interest, which is impacted by sample size.  Cochran suggests a pilot study or other preliminary information about the population distribution can be used. For official statistics with which I worked, we had previous sample and census data that could be used.  I have a question on ResearchGate asking others how they get the preliminary sigma (estimated square root of the population standard deviation). 

 https://www.researchgate.net/post/What_are_good_ways_to_estimate_sigma_for_sample_size_determination?_tpcectx=profile_questions

If you can at least estimate a "worst case" and "best case" for sigma, you can use it to estimate sample size requirements.  I have a paper on my profile page for estimating sample size from sigma for regression for finite populations, that turns out to be of the same format as for simple random sampling shown in Cochran, and referenced in my paper.

It gets more complicated if your population should be broken into strata, and whether or not you are estimating for statistics within those groups.  It is highly desirable to stratify if possible, but that depends upon your application.

An appropriate sample size 'formula' then depends upon many things: the kind of data, population, and goals.  The only simple such formula you are likely to see is for a worst case for yes/no data, for an infinite population without stratification.  Otherwise, you need to know more about your particular application.

Cheers - Jim

PS - Note that estimation using power of a test can be hard to interpret well. 

https://www.researchgate.net/publication/262971440_Practical_Interpretation_of_Hypothesis_Tests_-_letter_to_the_editor_-_TAS

Article FINITE POPULATION CORRECTION (fpc) FACTOR

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

Dear Hemal,

In my previous post i gave the summary of various ways. Here is a bit more detailed information regarding the same.  As stated earlier, many criteria are used to ascertain sample size such as types of scales used in measurement, availability of population size, use of pilot study results, or the possible scope of errors, and also the use of various analytical techniques. The sample size can be based on;

 1) Sample size table

Even tables are developed that help in easily finding out the sample size. Krejcie & Morgan (1970) have developed a table that helps the researcher determine (with 95 percent certainty) the sample size, and as per his table, a sample size of 384 is sufficient for a population size of more than 1000000.

2) Based on types of scales used in the instrument designing:

Based on the type of scale used for measuring variables, sample size can be ascertained. If study primarily uses 7-point likert scale, as per Cochran’s sample size formula for a seven point scale likert, sample size of 118 is deemed sufficient. 

3)     Based on number of factors to be analyzed: 

Literature (MacCallum, Widaman, Zhang, & Hong, 1999) reveals that for factor analysis, adequate sample size is partly determined by the strength of data, i.e. if the data strength is good, even for the smaller sample, results can be accurate and reliable. However, strength of data cannot be ascertained in advance, and thus it is important to have criteria for ascertaining the sample size.

In this regard, Comrey and Lee (1992) offered a rough rating scale for factor analysis and gave rough rating scale for sample size in factor analysis as: 100=poor; 200=fair; 300=good; 500=Very Good; and 1000 or more excellent. They urged the researchers to obtain samples of 500 or more observations whenever factor analysis was being used. 

The best method of determining the sample size for Factor Analysis is subject to item ratio, and in this regard, the very common method approved by a number of leading names (David Garson, 2008; Everitt, 1975; Everitt, 1975, Nunnally, 1978, Arrindell & van der Ende, 1985) is called the rule of 10, wherein, it is suggested that there should be at least 10 cases for each item in the instrument being used.

Thus, if we go by this assessment, an overall sample size of around 450 would be deemed sufficient as in totality we deal with around 45 items.

Hope it helps.

Hi Hemal, Please have a look on this attached file. I hope it helps...

Dear RG colleague,

Determining the sample sizes involve resource and statistical issues. Usually, researchers regard 100 participants as the minimum sample size when the population is large. However, In most studies the sample size is determined effectively by two factors: (1) the nature of data analysis proposed and (2) estimated response rate.

For example, if you plan to use a linear regression a sample size of 50+ 8K is required, where K is the number of predictors. Some researchers believes it is desirable to have at least 10 respondents for each item being tested in a factor analysis, Further, up to 300 responses is not unusual for Likert scale development according to other researchers.

Another method of calculating the required sample size is using the Power and Sample size program (www.power-analysis.com).

Regards,