I plan to collect data for my research through two questionnaires. The first questionnaire is for collecting data regarding the two independent variables from managers, and the second questionnaire is for collecting data about the dependent variable from customers (Customers' responses). So, the question now is how to handle this in the statistical analysis, given that the number of questionnaires collected from customers will be greater than that will be collected from managers (about 350 managers and 500 customers)? Also, is it ok if we have inequivalent numbers of questionnaires from managers and customers, or is it a must to insert the same number of responses for each variable in the statistical program (such as SPSS) to statistically examine the relationship between them?
Thank you so much in advance.
Best regards.
I believe that the number of collected surveys depends on what variables and what industry are you interested in. However, you can find formulas about the minimum sample size on the web.
this study could give you a good insight into the survey structure and the minimum sample size formula.
Article Modelling the time duration until the adoption of residentia...
It is not a 'must' to collect the same number of responses from two different populations. What is important is that the number of randomized responses is sufficient to capture a representative sample of each population. So fewer responses from the management population is probably appropriate given that the management population is likely smaller than the customer population. Are both populations normal in a Gaussian sense? You may need to address the characteristics of each population when evaluating whether or not your samples are representative.
Christopher C Kelly Thank you very much for your clarification. I would also like to ask how can I statically handle those two sets of questionnaires from two different populations in the statistical analysis to examine the effect of one variable, whose data has been collected from a population, on another variable whose data has been collected from another population?
If you use two completely separate samples, then there will be no way to calculate a correlation between your independent and dependent variables. You have to have the relevant variables reported by the same respondents before you calculate anything about their relationship.
It's fine for the numbers to be different. Ideally, you might weight the numbers of managers and customers to the total populations, but if they are selected with equal probability then you have self-weighting samples (if there is no nonresponse bias).
If you have other data that ties them together (e.g., these are across multiple retail stores, and you are able to identify the retail store for each respondent), then one approach is to create summary scores for the managers and customers at each retail store, and then look at whether those summary scores are correlated across stores.
Another option is to run a t-test for two independent samples. That would tell you, say, whether managers are more satisfied than customers, but not how one affects the other.
The most important thing is to consider the fact that statistical results, models and conclusion are valid only if the conditions of the data for using the statistical methods chosen are met. Data from assessments on scales are ordered categorical and such data could never be regarded as being quantitative.
Ordered categorical data have so-called rank-invariant properties, which permits monotonic transformations, only. Irrespective of the types of codes, ordered categorical data do not have standardized magnitude or equi-distance, which are the properties of quantitative data. There are methods for global scoring of multi-item instruments.There are also powerful statistical methods for statistical descriptions of ordered categorical data, for statistical modelling and tests.
Svensson E. Guidelines to statistical evaluation of data from ratings scales and questionnaires. Journal of Rehabilitation Medicine 2001;33: 47-48.
Svensson E. Construction of a single global scale for multi-item assessments of the same variable. Statistics in Medicine 2001; 20:3831-46.
Svensson E. Recent developments in analysis of paired ordinal data. Discussion of the paper Ivy Liu and Alan Agresti: The analysis of ordered categorical data: An overview and a survey of recent development. TEST 2005;14(1):1-73 pp 44-46.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Control Theory
- Control Systems