Discussions Promotions Jobs Members Search Aastha Kochar Follow Aastha Sample size determination Aastha Kochar Phd Research Scholar at Birla Institute of Management Technology Top Contributor If my questionnaire (for the purpose of writing a research paper) is having 33 statements (on likert scale), is there a minimum sample size that I should take or even a small sample of 40 respondents would do? Like (2) Comment (20) Follow Reply Privately 1 day ago Comments Hanlie Van Wyk, Muthoka Musyoka like this 20 comments Jump to most recent comment Wendy Appleby Wendy Wendy Appleby Analyst at a trade association Top Contributor For a start, what accuracy do you require? Like Reply privately Flag as inappropriate 1 day ago Aastha Kochar Aastha Aastha Kochar Phd Research Scholar at Birla Institute of Management Technology Top Contributor Mam, I want good accuracy and as per my knowledge reliability must be greater than .7 Like Reply privately Flag as inappropriate 1 day ago Jari Metsämuuronen Jari Jari Metsämuuronen Adjunct professor, Senior researcher / Evaluation Specialist at Finnish Education Evaluation Centre (FINEEC) Top Contributor yeps, there are very fine formulae for sample sizes and they usually require the info of how much error you can bear - the less population the more error in estimation - Wendy refers to that. The sample size depends on your population also - if you can find only 40 cases in your basic population, be it then. Usually, however, there are much more units in the total population than 40 only. There are also some quick and dirty rules for some methods. For Exploratory Factor analysis, for example, you might be willing to collect minimum 5 cases per item or 200 - 300 total (according to Comrey & Lee, 1992). Then 40 cases is not enough for this type of analysis. However, with 40 cases you can do lots of analysis but then you cannot use the 33 statements separately. If you have had some idea in the statements, that is, there is a theoretical background behind the statements, you can use that in calculating the sums for each diemension (without any EFA) and use the classical tools for reliability in assessing whether some of the items could be outselected from the sums. Item-total correlation is useful in this. Then, after summing up the relevant variables, you can do some univariate analysis like ANOVAs. The statistical inference urges, however, that you have quite clearly in your mind the population; the significance refers to how far you can generalize the result to the population. with nonrandom sampling the statistical testing is more or less questionable. Like (7) Reply privately Flag as inappropriate 1 day ago Towfique A., Antonia D. and 5 others like this Nancy Delosh Rusinak Nancy Nancy Delosh Rusinak HR Analyst/Statistician providing analytical support to find answers to workforce challenges What is your target population size? You also need to consider what you expect your participation and abandonment rates to be. There are plenty of sample size calculators online. They will allow you to adjust for error. But youll need to adjust the sample size to account for participation & abandonment. Like (3) Reply privately Flag as inappropriate 1 day ago Gene Shackman, Wendy Appleby and 1 other like this Gene Gene Shackman Director at The Global Social Change Research Project Right. As others have indicated, we need more info. 1. Whats the population size? 2. Can you do random or systematic sampling? 3. Any expectation of refusal rate (or abandonment rate)? 4. Do you want to make statements about each 33 statements individually or can you combine them? 5. Do you want to generalize to the total population or will you want to generalize to specific sub groups in the population? 6. What accuracy do you want, plus/minus 10%, 5%? Like (3) Reply privately Flag as inappropriate 1 day ago Gemechu Shale O., Jari Metsämuuronen and 1 other like this Aastha Kochar Aastha Aastha Kochar Phd Research Scholar at Birla Institute of Management Technology Top Contributor Thanks for your valuable input everyone. I am trying to write an empirical paper on Impact of CSR and corporate reputation on employee retention in Indian Service Sector (Indian banks on the basis of convenience) The research questions are: How CSR impacts employee retention? Does CSR impact corporate reputation? My questionnaire has 33 statements. Please guide me on the test that I should use in this and what should be the appropriate sample size. This is the first time I am writing an empirical paper so, things are not that clear. Like (1) Reply privately Flag as inappropriate 23 hours ago Gemechu Shale O. likes this Gene Gene Shackman Director at The Global Social Change Research Project So your population is Indian Banks? How many will you be getting data from? How many are you sending your questionnaire to? How many are there? What other variables will you have data about? Employee retention, location of the bank, size, age of the bank? Any info about profitability of the banks, their financial status, average salary of employees, whether the banks have any kind of retirement plans, other benefits for their employees? Like (2) Reply privately Flag as inappropriate 22 hours ago Hanlie V., Gemechu Shale O. like this Aastha Kochar Aastha Aastha Kochar Phd Research Scholar at Birla Institute of Management Technology Top Contributor Yes my population is Indian banks with a CSR reputation, which generally all pvt banks do have to some extend. I distributed the questionnaire to approximately 50 people as of now as per my convenience. The variables are CSR (Independent), Reputation (Dependent) and Retention (Dependent) Secondly, I dont need any information about the salary or financial status of the banks. The 33 statements (From Standard likert scale) are as below: CSR 1. My bank monitors new opportunities which can enhance banks financial strength 2. My bank allocate resources to improve long-term profitability 3. Bank comply immediately with new laws and court rulings 4. My bank is concerned with societys issues 5. My bank maintains a strong and competitive position 6. My bank always looks for new opportunities and programs which can improve community life 7. My bank respect ethical/moral norms adopted by society REPUTATION 8. I would be very happy to spend the rest of my career in this bank 9. I really feel as if this banks problems are my own 10. I do not feel like part of my family at this bank 11. I do not feel emotionally attached to this bank 12. My bank develops innovative products 13. It offers high quality products/ services 14. Bank has a strong record of profitability 15. It looks like a low risk investment RETENTION 16. They tend to outperform their competetors 17. They have maintained high standards in the way they treat people 18. I respect and trust my bank 19. I have good feelings about my bank 20. I am exposed to too high noise level 21. i have access to the necessary equipment to do my job properly 22. I am inconvenienced by the dust in my working environment 23. My work situation involves only a limited risk of industrial accidents 24. My supervisor looks for opportunities to praise positive employee performance both privately and in front of others 25. I feel undervalued by my supervisor 26. The supervisor almost never gives me feedback about how well I complete my work 27. My supervisor rewards a good idea by implementing it and giving the responsible employees credit 28. My chances for being promoted are good 29. There are enough career opportunities for me in this bank 30. Job vacancies are often filled by people from outside this bank 31. It would be easy to find a job in another bank 32. My bank prefers his employees to outsiders for any job in bank 33. An employees career development is important to this bank Like Reply privately Flag as inappropriate 22 hours ago Gene Gene Shackman Director at The Global Social Change Research Project Well, 2 points. 1. You wrote I distributed the questionnaire to approximately 50 people as of now as per my convenience.How did you choose the 50 people? Does per my convenience mean they were people you knew, or they were at banks you know? That is, they were not a random sample? If the people were not randomly or systematically selected, then you really cant determine the impact of CSR on employee retention. You dont know in what ways the sample may be biased. Are these 50 people from 50 different banks? 2. Also, you really do need additional variables, any you can get. There may be other explanations for employee retention, that you wont know about unless you collect the data. For example, if some banks pay more, or have retirement plans, while others dont, and those banks with plans have higher retention, then the retention is due to higher pay or retirement plans. 3. What you can do is to report correlations of those questionnaire responses to employee retention. You cannot establish causation, but you can describe correlation. You can also look at any other literature you can find on employee retention and see whether your research fits with the results of those other studies. If your results are consistent, you are better able to say that your research may be consistent with the idea that CSR is related to employee retention. Hope that helps. Gene Like (5) Reply privately Flag as inappropriate 21 hours ago Ruth H., Jari Metsämuuronen and 3 others like this Aastha Kochar Aastha Aastha Kochar Phd Research Scholar at Birla Institute of Management Technology Top Contributor This was a great input from your end sir. Those 50 people are those in my circle and who r working in banks having a CSR reputation. Moreover, these are people from different banks and not any one particular bank.Like I gave the questionnaire to my friends and asked them to get it filled from their collegues at the bank, this is how data has been collected due to time constraint. Frankly, for the time being I need to submit an empirical paper and I just dont want tosubmit, but be justifiable in whatever I report. I came across literature that proved that CSR do impact employee retention and these statements are adopted from one such paper only as a standard scale. Any other inputs are most welcome. Do I need to distribute the sample to more people in order to apply further test? Please guide me on this. Like Reply privately Flag as inappropriate 21 hours ago Gene Gene Shackman Director at The Global Social Change Research Project You can write a paper, and show correlation. You cant claim causation, but only correlation. You can say your results are consistent with at least one other paper. If you can find more research that would be good. Also, if you can get any kind of random sample, from more people, that would be even better. Like (1) Reply privately Flag as inappropriate 21 hours ago Gemechu Shale O. likes this Jari Metsämuuronen Jari Jari Metsämuuronen Adjunct professor, Senior researcher / Evaluation Specialist at Finnish Education Evaluation Centre (FINEEC) Top Contributor Your expression I need to submit an empirical paper and ... be justifiable in whatever I report makes me think that you can choose from two paths. First, you report whatsoever and then there is not much need for collecting more data or thinking the randomness of the sample. This will be a report of kind of report something and we do not care the result - it is done just to show that we do research. This may be relevant option if you, in Birla institute, need to show any/some activities three or four times a year in order to get some governmental support/tax reduces or such (in Nepal there is this kind of system). This is not an option for a real research(er). The second option is for the case that you are willing to do serious research. Then you need to think twice your sampling frame and population. What is the population you are willing to generalize the results? Whole India - or Uttar Pradesh State - or Greater Noida town area (I took these from the address of your institute)? Are there some other characteristics in the banks which may be relevant when selecting them - like the size of the bank or level of international activities? If so, then these should be part of your sampling frame. In that case, you, most probably, are willing to go to some kind of stratified sampling strategy. If all the banks are at the same line, you may select a strict random sampling strategy. This option leads you to the ideas discussed by Gene and other colleagues. And one more: the error, discussed by Wendy and other, has not much to do with the reliability of your instrument - that is another discussion. Its good to have reliability higher than .70, as you wrote. The error here refers to the error in your output result. With 5-point Liker scale (0 - 4) and 7 items, the range in your DV is 0 - 28 points. How much error you can bear in the final score? 5% (1,4 point plus minus), 10% (2,8 points plus minus), 20% (5,6 points plus minus)? Like (4) Reply privately Flag as inappropriate 21 hours ago zohreh Shahghasemi, Darrin Thomas, PhD and 2 others like this Aastha Kochar Aastha Aastha Kochar Phd Research Scholar at Birla Institute of Management Technology Top Contributor I understand your point, but for now I dont have much time to plan the sample using stratified sampling. Moreover I will gain insight on your input from my guide as well. Thanks a lot Like (1) Reply privately Flag as inappropriate 20 hours ago Gemechu Shale O. likes this Aastha Kochar Aastha Aastha Kochar Phd Research Scholar at Birla Institute of Management Technology Top Contributor Moreover I am not doing this for any rebate or other benefit. Its just that I dont have expert guidance and secondly this came on a very short notice. Like (1) Reply privately Flag as inappropriate 20 hours ago Gemechu Shale O. likes this Caleb Ouma Caleb Caleb Ouma Senior Socio-Economist at GIBB International Ltd Kochar, do not use shortcut in research, it may cause you a lot of embarrassment in the future. They best thing to do is to establish a target population that you can use. This will help you when analyzing your data and generalizing your result. Like (3) Reply privately Flag as inappropriate 14 hours ago Stelios K., Jari Metsämuuronen and 1 other like this Aastha Kochar Aastha Aastha Kochar Phd Research Scholar at Birla Institute of Management Technology Top Contributor Thanku everyone. Like (1) Reply privately Flag as inappropriate 14 hours ago Gemechu Shale O. likes this Stelios Katsoulis Stelios Stelios Katsoulis ICT primary school Teacher at Ministry of Education, Lifelong Learning and Religious Affairs As you probably read above, your target population will determine your sample size. Yet even if you successfully conduct reliability analysis, with a N=40, you will have an issue, when you will try to conduct a factor (or a PC) analysis. It may not sound as necessary, but it will become, when you will try to prove that the measuring instrument (the questionnaire in other words) truly measures what you designed it to. Like (2) Reply privately Flag as inappropriate 12 hours ago Gemechu Shale O., Wendy Appleby like this Yohannes Asmare Chekole Yohannes Asmare Yohannes Asmare Chekole Senior Lecturer at Bahir Dar University, Ph.D. Candidate in Social Psychology at Addis Ababa University Determination of size of samples requires consideration of lots of issues:the population, nature of the research design, resources, nature of the problem and so on. Like (2) Reply privately Flag as inappropriate 11 hours ago Gemechu Shale O., Wendy Appleby like this Kenneth Kenneth Mutai Data Manager at University of Nairobi The following steps might be of help: 1. CSR, Reputation and Retention will be summarized as average scores 2. Choose the kind of statistical test you will use to test the relationships between independent and dependent variables; in your case it will be Pearson correlation test - coefficient (r). 3. Choose your null hypothesis (H0): it could be r=0 (no relationship) or r=0.5 (below this the relationship is weak) 4. Choose the alternate hypothesis (H1); desired r: it could be 0.7 or more. Note: The higher the difference between the H0 and H1, the smaller the sample size required and vice versa. 5. Choose your desired power and alpha. 6. Finally, subject the above to an appropriate formula or use statistical software such as PASS or StatsDirect. You get your sample size. Then the issues of sampling strategies as outlined above by various commentators will apply. Hope this will give you some insight. Like (1) Reply privately Flag as inappropriate 9 hours ago Hanlie V. likes this Aastha Kochar Aastha Aastha Kochar Phd Research Scholar at Birla Institute of Management Technology Top Contributor Thank u for your valuable input Kenneth :) I will surely work in this manner Like Reply privately Flag as inappropriate 9 hours ago
Posted on: Fri, 29 Aug 2014 01:36:28 +0000