Search

Objective: In clinical studies, the relationship between dependent (Y) variable and independent (X) variables is expressed by regression analysis. However, if one or more of these variables contains uncertainty and outliers, classical regression analysis cannot be performed. The application of the fuzzy linear regression analysis approach was introduced when there were ambiguous and outliers in X variables and it was discussed that more reliable estimates were obtained on a clinical sample. Material and Methods: We showed whether there is a statistically significant difference between observed values and estimated values calculated using fuzzy linear regression analysis methods based on classical and linear programming. In clinical studies, we suggested an h=0.5 value, which we call "turbidity tolerance level". We used mean squared error, square root of mean squares error and the coefficient of determination (R2) indexes as the goodness of fit test criteria showing the compatibility between the values calculated at the suggested h-level. Then, we showed and interpret the calculated values graphically. Results: With classical and fuzzy linear regression analysis methods, standard error and R2 values of the estimated average weights of newborns were calculated as 2.635 (g) ± 32.82 (g);R²observed/Kestimated=0.61 and 3.117.72 (g) ± 21.97 (g); R²observed/Bestimated=0.97, respectively. The turbidity of the fuzzy linear regression model created in 22 iterations was calculated as 49.789. Conclusion: According to the approach, the effect of the mother's age, birth weight, education level and number of fasting days on birth weight was found to be significant. It has been suggested that the method can be used in clinical studies as an alternative approach. [ABSTRACT FROM AUTHOR]