What do you intend to do with the data? To generalize to a larger population? Do you have a random sample?

In any case, gpower (http://www.gpower.hhu.de/en.html) is a good way to help you decide what target sample size you need to achieve for your purposes.

Before you are going to collect the data, you should think of the minimum sample size based on your methodology. Then after, you can use those methods such as online, survey....to collect the predetermined sample. 

The sample size that you need depends on the type of analysis that you wish to do.  For example, if you wish to do Structural Equation Modelling, you need a sample of at least 200 respondents.  You also need to check that your target market is adequately represented by those people who have responded to your invitation.

Also worth knowing when you need the project done by. Online methods can allow you to collect much more data that might be suggested by a power analysis (and you might not know how small you want your standard errors). If that is the case you might work backwards from the deadline and how long it will take you to write up the report. There are lots of different ``stopping rules,'' and while power can suggest one others also exist.

in addition to the information above your sample size depends on how many variable your study has ?

also you can take these points in account in order to determine the sample size:

First: the size of the appropriate sample depends on the purpose of the study , and the nature of the research community in addition to the variables of study, and the pattern of relations that he wishes to disclose.

Second, it is possible to infer the appropriate sample size on previous studies, if any, especially those studies that have the same research design.

Thirdly, the increase in sample size can provide a higher representation of the characteristics of the community and thus a more accurate dissemination of the results of the research. Other factors affect the size of the sample

The nature of the approved study methodology

It all depends on type of analysis you are going for and how minimum do you want your standard errors. Other than that there many formula's such as Green's or Cohen's that can actually help you out.

Sometime you don't know your population size then there are some other as well that can help you out.

A general rule of thumb (Bryman) is to go for 300 or plus however different researchers have answered it differently.

as a thump rule 384 is considered to be enough sample size.However ,depending upon the variables you can multiply by 7 to determine the sample size

thank you for the information you posted here.the responses I get are more than 384. Can I call the sample random or accidental or appropriate what what kind of sample should  I state in the methodology?

It is random sampling procedure as you are not sure who will respond to your mail .

It appears to be a general survey with no particular geographical region in mind. When an online survey is available online, then anyone could participate. There is even a risk of multiple responses by same person, since it was not by invitation. As members have alluded, the size depends on your research objectives.

In our case, we will invite specific people (by using SurveyMonkey tool) to respond to an online survey and not the general public.

I desagree with some of the previos answers. Your sampling scheme is not random at all, probabilities are unknown a priori. Your respondents decide to answer the survey, so you can't generalize your results beyond the sample. If you have more than 400 cases and the structure of the sample in control variables is similar enough to the population ones, you can make some inference.

384 ( derived using yamane formula,) is appropriate if you know the size of the population is known 

First of all, when you like to post your questionnaire on social media,  it is very unlikely that you will get hundred of responses quickly. Moreover, statistical significance can be ensured if you use random sampling to select respondents.

Hence, you can decide which social media platforms you like to use for survey. Many studies used facebook groups to select  online respondents. I feel in your case you can use online sampling method suggested by Malhotra and Das (2014)

Here is a case example I like to mention based on approach used in few journal articles

1) Use purposive sampling to select online social media groups which represent your target audiences.

2) Use Systemic sampling or random sampling to select respondents from each group.

Sample size

As others researchers already

indicated that sample size depends on research method, number of variables, model complexities, etc.

1)For example, for regression analysis sample size depends on number of predictor variables.

2) For SEM based analysis, sample size (N) = number of observe variable (n) * 20. If there are 30 observe variables, sample size would be 300. Some authors suggested N= n* 10 sample size.

Sorry, in my previous post , please read  sample size would be 600 (30*20).

This is a survey with unknown biases: it is not clear who would see the invitation and who would respond. That makes it hard to project the results on the general population. Whether and how much that matters depends on several things. What is the target population that you wish to describe? How well do the chosen media reach that population? Will you collect enough data about the respondents so that you can weight the results to better match the target population? How much are the results likely to vary depending on who responds?  Any write-up of your results should address these issues so that people know how to evaluate your results.