The survey sample of 1,635 was selected in systematic, stratified fashion by KMPS and Readex Research from segments of MD&DI’s domestic circulation in the categories of finished medical device manufacturers and/or in vitro diagnostic manufacturers, representing 33,287 recipients at the time of sample selection.
Data was collected via mail survey from August 21 to October 2, 2008. The survey was closed for tabulation with 542 usable responses—a 33% response rate. As with any research, the results should be interpreted with the potential of non-response bias in mind. It is unknown how those who responded to the survey may be different from those who did not respond. In general, the higher the response rate, the lower the probability of estimation errors due to non-response and thus, the more stable the results. The margin of error for percentages based on smaller sample sizes will be larger.
The final results are based on the 460 individuals who indicated they are involved in the industry and work full time in firms manufacturing finished medical devices and/or in vitro diagnostics, representing approximately 28,000 MD&DI recipients. The margin of error for percentages based on 460 usable responses is ±4.5% at the 95% confidence level. The margin of error for percentages based on smaller sample sizes will be larger.
The analysis of the Medical Device & Diagnostic Industry Salary Survey data used multiple regression analysis to model the determinants of salary by identifying those variables which, when taken together with appropriate weights, provide the best prediction of any individual’s actual salary.
The final salary prediction model is somewhat restricted in its applicability—it represents only full-time professionals (employed year round) who are under 65 years old with salaries in the range of $34,000 to $240,000. Only those whose salary is at least 60% of their compensation were considered.
Statistically speaking, this model is moderately powerful: it explains 53% of the variation in salary (adjusted R-square = .533), and is significant by the F-test at p<.000.
While a model explaining about 53% of the dependent variable’s variation may be described as "moderately powerful," it still leaves about half of the variation unexplained. It is virtually certain that other variables not captured through this survey also have an effect on salary levels: individual job performance, for example. To the extent that this model does not include variables actually important in determining salary, its conclusions must be interpreted cautiously.
