| Purpose: To evaluate the application of experiment design schemes commonly used in medical thesis of 2000,2001and 2002.To analysize the purpose, treatment, non-treatment factor and interaction of experiment design so as to find out the incorrectness or impropriety and arose the alert of research staff and close attention of administrative departments and give some advise to improve the quality of medical thesis. Methods: According to the terms that whether accord with experiment purpose, amount of treatment variables, interaction between treatment variables and balance conditions of non-treatment variables, divide the application of experiment design schemes, including grouped design, paired design, randomized block design, cross-over design, Latin squares design, factorial design, orthogonal design, repeated measure design, sequential design, into four analysis result indices, i.e. none, incorrect, ambiguous and correct. Altogether 657 thesis, in which 293 of 2000, 183 of 2001 and 181 of 2002, were sampled from the library of Hebei Medical University and the library of Hebei Province. The experiment design schemes used in these theses were analyzed one by one and sample clustering was done. Total distribution comparison and comparison in couples of ordinal data were done by using Kruskat-Wallis H test and Nemenyi test; comparison of the constituent ratio of the four indices, none, incorrect, ambiguous, correct, in each category was tested using χ2; comparison of the distribution of single factor design and multi-factor design was analyzed using Ridit, the distribution comparison of those indices in the two design models was tested using χ2; comparison of the application of the experiment design schemes in recent 3 years and the further comparison in couples were analyzed using Ridit analysis; the constituent ratio of single factor experiment design and multi-factor experiment design in the recent 3 years was tested using χ2, the further comparison in couples was tested with corrected α. Results: Among the evaluated 657 thesis, the proportion of none, incorrect, ambiguous and correct are 6.4%, 11.6%, 18.7% and 63.3% espectively, 36.7% of which have something wrong in the use of experiment design. The problems mainly are: 1) the experiment design doesn't follow the principle of randomization; 2) incomplete comparison setting and repeated comparison; 3) short of comparison; 4) neglect of the interfere of confounding factors; 5) no perfect experiment design scheme, improper selection of experiment design method or the reasoning of the design method is not enough; 6) experiment design scheme doesn't match the experiment purpose; 7) replace multi-factor design with single factor design; 8) incomplete assembly of experiment factors in various level. The commonly used experiment design schemes fall into 3 categories, category I includes factorial design, orthogonal design and cross-over design; category II includes paired design, Latin squares design, repeated measure design and sequential design; category III includes grouped design and randomized block design; with significant difference between each category(HC=26.8071, p=0.0000). The application of multi-factor experiment design scheme and single experiment design scheme was compared with statistical significance in their difference (u=2.1236, p=0.0337). The indices such as incorrect of single experiment design scheme and multi-factor experiment design scheme has significant difference (χ2=12.6250,p=0.0004). The indices such as none, ambiguous and correct all have no significant difference (p>0.05). The distribution difference in the 3 categories of the experiment design schemes in the medical thesis of recent 3 years is very significant (χ2=44.0895, p=0.0000). The experiment design schemes of 2002 are better than other two years. Conclusion: In recent 3 years, the application of experiment design schemes in medical thesis has serious problems and should arouse enough attention. All 3 categories have certain problems, the applicatio...
|
PDF Full Text Request