Is Sample size of 200 nos , sufficient for research conducted for development of new scale .? that is what i have planned ?
i have been working on a research for Development of education system at basic level which builds a strong education attitude with innovation and help me develop a new scale to measure the strength of education system core components .
One guide will be to determine the sample size by the number of items from your questionnaire by a factor 10. Say you have a 10 item questionnaire, then when multiplied by 10, it gives you 100, so you need at least 100 persons.
For the development of scale, you are required to test validity of scale. For the same, you need to perform CFA. For CFA, minimum sample size requirement is 15 to 20 times number of items in the scale. You should have at least 300 sample size.
Typically, the rough rule of thumb in test construction has been a minimum of 400 males and 400 females (which was possibly recommended by Anne Anastasi many years ago... but not sure). However, I would stress that this is the absolute minimum! Additionally, you should aim for a socioeconomic and ethnic background in your norming sample that mirrors the most recent population census. Therefore, you will need much more than this minimum to norm (i.e., standardize) a psychometric instrument for use in a particular population! Otherwise, your test will be 'normed' on only some sub-sample of the population.
Interesting, I am involved in the development of a CIR Measurement Tool with an EAP Vendor and we just submitted to the IRB.... Our statistician who has created similar scales claims that a sample of 200 is sufficient for the Beta Test to check for validity and reliability.... perhaps it depends on the field of study...
I will suggest that the suggestions of Partricia and Tali above should be substantiated with references so that you can use that to support the choice of your sample size but there is also G*Power software that can be used to establish sufficient sample size. It might be what the statistician mentioned above used to make the calculation. G*Power is a free statistical tool that can be downloaded from online.
I have previously stated that your assessment instrument would need to be normed on “a minimum of 800” adults (400 males and 400 females), and that this would be an “absolute minimum” (with some degree of humour in stating this). To be more specific, 800 would be a “minimum number” in a very homogeneous populations such as in Japan. In a much less homogeneous populations however, such as India, it would be utterly impossible to adequately norm a psychometric instrument on 800 let alone 200 individuals as I have heard in this discussion! It is a mathematical impossibility that 200, or even 2000 individuals, could accurately represent the complexity of India’s very large and very diverse population of 1,300,000,000 people, with more than 29 different ethnic and language groups. In addition to these very diverse and often uniquely different ethnic/language groups, a test developer would need to take into account very diverse and different religious differences as well (approximately 80% Hindus, 14% Muslims and the remaining 6% representing more than 4 other religious groups); and if one was able to adequately represent that ethnic and religious complexity, the ‘easy part’ would to be represent a fairly equal gender, educational and socioeconomic breakdown of the country! The numbers can no doubt be worked out that you would need taking into the basic population breakdown presented above, but I assure you that it would be much more that what has been suggested in this discussion so far! If one thinks that this can be logically and realistically accomplished with 200 people taking into account the reality of India’s population complexity (only briefly described above), please feel free to challenge my thinking. I have developed 12 psychometric instruments since 1978 (one of which has been normed in many countries around the world), and I assure you that no serious test publisher in the world would publish any instrument normed on 200 individuals in such a massive and diverse population as India.
I agree with Reuven. Sample size is influenced by many factors. One issue that a researcher must consider is the size of the parent population. There is long-standing research that roughly 1000 participants can generalize to a population over hundred thousand. Another important factor, is the sampling procedure that one chooses to use. I recommend stratified random sampling for such cases. This allows you to match your sample's competition with the parents population. Also, consider the fact that smaller sample sizes are more sensitive to random error. Taking all these things in mind, I certainly would be uncomfortable with a 200 participant validation sample.
The best response to your question is to conduct a G*Power Analysis, which will provide the most accurate sample size for your study. This program is "Free" and is offered through the Department of Psychology, Germany. The developers of the program are added below for your perusal:
Department of Psychology
G*Power has been developed by these people:Dr. Franz Faul, Department of Psychology, Christian-Albrechts-University, Olshausenstr. 62, 24098 Kiel, Germany. Dr. Edgar Erdfelder, Professor of Psychology, Department of Psychology, Mannheim University, Schloss, Ehrenhof-Ost68131 MannheimGermanyDipl.-Psych. SAE-Diplom Audio Engineer Albert-Georg Lang, Department of Experimental Psychology, Heinrich-Heine-University, 40225 Dsseldorf, Germany. Dr. Axel Buchner, Professor of Psychology, Department of Experimental Psychology, Heinrich-Heine-University, 40225 Dsseldorf, Germany.
Innovationsprozesse in frühen Phasen werden beschrieben durch Adjektive wie „unstrukturiert“, „fuzzy“, „emergent“, „firmenübergreifend“
und „kooperativ“. Oft ist nicht ganz eindeutig, wer an einer Innovation beteiligt ist und wie der Beitrag einzelner Akteure
zu bewerten ist. Fluide Teilnehmerstrukturen und schwer zurechenbare Leistungen setzen der...