 |
 |
Dear colleagues I need some help in interpreting some STATA output which is quoting AIC/BIC. Effectively, I am using MIDAS to pool some sensitivity and specificity values for different cutoffs on a diagnostic test. Alas sensitivity and specificity may vary not only dependent on the cutoff but also due to variation in the background vaccination rate for a related condition. I have generated the following output from midas, which quotes BIC and AIC for the contribution of the variable 'vaccrate'. Could anyone help me interpret whether this could be considered as a significant contribution or not? Estimating Covariate Effect Of: vaccrate Sensitivity and Specificity Joint Model SUMMARY DATA AND PERFORMANCE ESTIMATES Number of studies = 18 Reference-positive Units = 6185 Reference-negative Units = 625 Pretest Prob of Disease = 0.91 Deviance = 263.5 AIC = 273.5 BIC = 281.4 BICdiff = 657.1 Correlation (Mixed Model)= -0.90 Proportion of heterogeneity likely due to threshold effect = 0.80 Interstudy variation in Sensitivity: ICC_SEN = 0.17, 95% CI = [ 0.05- 0.29] Interstudy variation in Sensitivity: MED_SEN = 0.68, 95% CI = [ 0.62- 0.77] Interstudy variation in Specificity: ICC_SPE = 0.20, 95% CI = [ 0.09- 0.31] Interstudy variation in Specificity: MED_SPE = 0.70, 95% CI = [ 0.65- 0.77] ROC Area, AUROC = 0.84 [0.80 - 0.87] Heterogeneity (Chi-square): LRT_Q = 246.886, df =2.00, LRT_p =0.000 Inconsistency (I-square): LRT_I2 = 99, 95% CI = [ 99-100] Parameter Estimate 95% CI Sensitivity 0.69 [ 0.59, 0.78] Specificity 0.84 [ 0.77, 0.89] Positive Likelihood Ratio 4.3 [ 3.3, 5.7] Negative Likelihood Ratio 0.37 [ 0.29, 0.47] Diagnostic Odds Ratio 12 [ 9, 16] Thanks to all Neville You may leave the list at any time by sending the command SIGNOFF allstat to [log in to unmask], leaving the subject line blank.