Hi Shaker, it is not good practice to draw a sample and then ask others to tell you what you actually did. I hope you employed more care in your questionnaire and data analysis. According to what you write you employed cluster sampling, i.e. randomly selected a cluster of hotels and then surveyed all department managers in these hotels. From what you write I don't know whether 231 managers was the right sample size (and I guess you don't know either). I'd certainly not use a table from 1970, but rather try to find out how the key variables of interest vary across the population of department heads and across hotel clusters.

How are you collecting your data? I ask because, for inexpensive methods of data collection (e.g., online or paper survey), it might be practical to  collect data from all 583 department heads.

Hi Carl, by the way i used online survey ( monkey survey) and here what i justified in data collection :

The data collection process done via an electronic-mail connected to web address (https://www.surveymonkey.com/s/LTVHCJJ).This method was used for the survey and sent to HR department in every hotel, where in turn every HR department in each hotel distributed the survey through emails to the respective departments. Data collection was conducted from the mid of July 2012 to 25 of august 2012. The survey questionnaire attached to the e-mail was with cover letter which stated the purpose of the study for Jordanian economy to be answered by the head of the department (Appendix A).

At the end of July 2012 only 154 questionnaires answers were collected, constituting a response rate of 65.8%. Since the rate of response was low, the researcher adopted another strategy in the beginning of august 2012 by calling directly the HR manager via phones in order to get their cooperation to distribute the survey through emails to the required departments.

The researcher managed to obtain another 77 responses after phone calls. This was done by calling the HR managers directly to check whether they have received the questionnaires or not. Once they have confirmed receiving the questionnaires, the HR managers were kindly requested to distribute the survey through email to all the departments head.

i need to know what is suitable sampling method ( technique) to my study? since the number of departments head already known in each hotel ( 11 HODs). only i used random sampling to select hotels that can cover the sample size ( 21 hotels out of 53 * 11 department in each hotel = 231 ( sample size).

Just to clarify, are you asking about the technical term for the sampling method that you employed?

there are numbers of sample  techniques to choose sample from a population such as  simple random sampling, stratified random

sampling, cluster sampling, and systematic sampling . so i want know which one can applied to be fit  my study based on mentioned above.

Dear Shaker, why don't collect data from all 583 department heads. You just send them all and wait their respected responses.

Dear Ra'ed , the data has been collected already. now i need to know  (based on what i have mentioned above), what sample technique is the most fit to my study : i.e ....random sampling or stratified random

sampling or  cluster sampling or systematic sampling .

With regards to this: 

"there are numbers of sample techniques to choose sample from a population such as simple random sampling, stratified random

sampling, cluster sampling, and systematic sampling . so i want know which one can applied to be fit my study based on mentioned above."

I dont think that any of those sampling methods fit--sampling decisions are typically made prior to data collection and it sounds as though you originally attempted to contact all of the HODs. Is that correct? If so, I would just say that you randomly selected (N=53) 4-5 star hotels and attempted to obtain survey responses from each department head, then give your overall response rate, and response rates for both online and phone surveys. To address non-response, you could try to obtain important characteristics of all 583 department heads in order to describe any differences between those HODs who responded and those who did not (for example, are there departments whose HODs were more likely to respond? Age differences? Length of time with company?). 

Dear Carl , thank you much , you have confirmed what was in my thought . the question I addressed above was the examiner question . anyway thank you :)

Hi Shaker, it is not good practice to draw a sample and then ask others to tell you what you actually did. I hope you employed more care in your questionnaire and data analysis. According to what you write you employed cluster sampling, i.e. randomly selected a cluster of hotels and then surveyed all department managers in these hotels. From what you write I don't know whether 231 managers was the right sample size (and I guess you don't know either). I'd certainly not use a table from 1970, but rather try to find out how the key variables of interest vary across the population of department heads and across hotel clusters.

Dear Franz Barjak,

Thanks for your comment. let me clarify more in the following points :

First : i agree with you about " it is not good practice to draw a sample and then ask others to tell you what you actually did". my respond to this point is that i am still in learning process ( Phd student), and i asked the question to clarify since sometimes sample technique make the researcher confuse.

Second : i dont even think that cluster sample is suitable to my study because i didnt cluster the hotels to four and five stars hotels. i just select randomly the hotels  ( dont mind if they were all five stars or 4 stars as long as they was in range 4 or 5 stars ).Therefore, i think the most suitable sample technique to my study is stratified sample for choosing the respondents and random sample for choosing hotels. the reason of choosing stratified sample because it used when there is a Homogeneity ( i.e all are head of departments) and heterogeneity( i.e the respondents were from different department). how do you think ??

Finally according to the table dear, i have used different sample method to choose the sample size ( i didnt mentioned here ) and the table mentioned was the most appropriate because it cover the most number of respondents between other methods used in the previous studies.

So what i need from the question above to see different ppl answers to confirm my thought.

thank you Franz

SHAKER

I am such worried on how you went to collected data without defining your sample and sampling techniques first if you are a PhD student and you want someone to do that for you. What case study strategy are you using (single embedded or multiple embedded). If your using single embedded approach, it would be better to use the stratified sampling method in your data collection since you are trying to solve a problem but using two similar organisation,

Franz, your comment is like the one we get from journal article reviewers "study design is inappropriate". I don't see any use of such comments when things are done. 

Look for the journal articles and see if your method fits your what you want.

Shaker -

I was wondering if you did a random sample with n = 21, and N = 53, where each survey had eleven sets of questions, which would lead us to think it was simple random sampling and if you were estimating something like mean expenditures for some item, then the estimate and it's standard error for each item would be found in any survey statistics text, such as Cochran's Sampling Techniques, 3rd ed, 1977, John Wiley & Sons. However, if your procedure, as Carl apparently figured out, was not a simple random sample, but, I guess, an attempt at a census, then the characteristics that Carl described could be used to derive "response propensity" groups, which you can research on the Internet. That would give you a method for forming these groups, something akin to poststratification in procedure. This would help you to artificially fill in your population, item by item. However, I would be worried about accuracy.

I understand that the data are already collected, and it sounds like this was brought to you after-the-fact. But for future consideration, you could note to whomever it concerns that, as stated previously, a survey does need to be planned all the way through, ahead of time. And when it does come time to assess results, there are both nonsampling errors to consider, such as frame errors and measurement errors, and if there is a sample rather than a census, there is sampling error to consider. Often total survey error may be lower for a sample, as a census can have a great deal more measurement and other nonsampling error. Here it sounds like a small population, but one that is not required to respond, as opposed to an official survey, so trying hard to obtain a random sample with little nonresponse might be better next time. Maybe. Cochran has a way to estimate simple random sample size needed, discussed on pages 77 and 78. Information from the current survey might be helpful there in the part he says you have to "guess." Usually stratification helps, but your n and N are both small, though you might stratify by number of stars for the hotel. Auxiliary data for model-assisted design-based sampling would be much better. Perhaps administrative data would be available for that. But to return to your current plight, it sounds like response propensity groups are the way to go for a census with a good deal of nonresponse. With voluntary surveys, this is a technique you may always need or at least consider.

Cheers - Jim

Shaker - 

Oops. I missed a few comments when looking at this on my phone, and I see that you are saying you collected randomly within strata?  If you are sure that was done, then you have your answer as to the name of the procedure, "stratified random sampling," but I am not certain that was actually what was done. Regardless, a book such as Cochran's - mentioned above - could be very useful, as well as response propensity type 'stratification.' 

Jim 

Dear James , thank you very much

Shaker -

One last thing you could try (which I have found very useful for highly skewed establishment surveys, and I was chatting with someone in, I think, England not very long ago who even found it very useful for household surveys), would be strictly model-based sampling and 'prediction.' This is not forecasting (as it may sound to be from the name), but uses the relationship between your sample data and regressor data on the population. Suppose you had some administrative data, such as number of employees for every hotel. Perhaps that would be a good regressor (x data) for a number of your numerical-based questions (y data). Expenditures for example. Then if you group your data (stratify) well, you can predict data for every missing y-value case out-of-sample. This can also be used to estimate (predict) for nonresponse on a census.

As in the case of design-based (randomized) sampling, the same sample size will do better for some questions than for others, and you will have to adjust according to your resources and the questions for which you most want accurate answers.

If this handles any of your requirements, then you might want to see the paper at the attached link.

Hope this helps.

Jim

Conference Paper Projected Variance for the Model-based Classical Ratio Estim...

Dear Researcher

In order to answer this question/problem, several remarks have to be studied.

1.

Research studies are usually carried out on sample of subjects rather than whole populations. The most challenging aspect of fieldwork is drawing a random sample from the target population to which the results of the study would be generalized. The key to a good sample is that it has to be typical of the population from which it is drawn. When the information from a sample is not typical of that in the population in a systematic way, we say that error has occurred. In actual practice, the task is so difficult that several types of errors, i.e. sampling error, non-sampling error, Response error, Processing error,…

In addition, the most important error is the Sampling error, which is statistically defined as the error caused by observing a sample instead of the whole population. The underlying principle that must be followed if we are to have any hope of making inferences from a sample to a population is that the sample be representative of that population. A key way of achieving this is through the use of “randomization”. There several types of random samples, Some of which are: Simple Random Sampling, Stratified Random Sampling, Double-stage Random Sampling... Moreover, the most important sample is the simple random sample which is a sample selected in such a way that every possible sample of the same size is equally likely to be chosen. In order to reduce the sampling error, the simple random sample technique and a large sample size have to be developed.

2.

The following factors are highly affected the sample size and need to be identified:

• Population Size,

• Margin of Error,

• Confidence Level (level of significance) and

• Standard of Deviation.

Then, the sample size can be estimated by,

Necessary Sample Size = (z-score or t-value)2 * StdDev*(1-StdDev) / (margin of error)2 .