All aspects of development in Analysis (which does not fit in another project)
FS#122  significance tests for an adjusted odds ratio (options /t and /ex)
Attached to Project:
Analysis
Opened by Tom Nichols (tom.nichols)  Tuesday, 10 December 2013, 17:43 GMT
Opened by Tom Nichols (tom.nichols)  Tuesday, 10 December 2013, 17:43 GMT

Details* EPIDATA ANALYSIS PROGRAM
logopen "H:\Teaching Courses\Outbreak Course\bugTN_06122013\bugTN_06122013.htm" /close /replace cd "H:\Teaching Courses\Outbreak Course\bugTN_06122013" read "H:\Teaching Courses\Outbreak Course\bugTN_06122013\wedd" /close set random simulations=10000 *********************************************** * EpiData Analysis 2.2.1.171 * tables ill redsauce gheerchick /o /nt /nc /ex /t /obs * Using 2.2.1.171: OR = 1.08 [95% CI 0.22 to 5.66]; p(MH)=0.916; p(exact)=0.0414 * EpiData Analysis 2.2.2.182 tables ill redsauce gheerchick /o /nt /nc /ex /t /obs /oex * Using 2.2.2.182: OR(Exact) = 1.08 [95% CI 0.22 to 5.66]; p(MH)=0.046; p(exact)=0.0414 * Using 2.2.2.182: OR = 1.07 [95% CI 0.11 to 10.06] * Epi Info for DOS v6.04d * tables redsauce ill gheerchick * Using Epi Info for DOS v6.04d: OR (MLE) = 1.08 [95% CI 0.22 to 5.66]; p(MH uncorrected)=0.91644269; p(MH corrected)=0.81105112 * Stata/SE 12.1 for Windows (32 bit) Revision 20 Mar 2013 * logit ill redsauce gheerchick * Pvalue for coefficient of redsauce = 0.916 * The estimate is close to 1 and the confidence interval is wide, so a test for the adjusted OR being significantly different to 1 should have a pvalue close to 1 * Both Epi Info for DOS and Stata give a pvalue of 0.916 * So I have a lot of confidence in this being the correct pvalue. * EpiData Analysis (build 171) * The chisquared test statistic and the pvalue for a MH weighted odds ratio using the option /t is 0.01 and 0.916 * This pvalue using the option /t appears to be correct. * The pvalue using the option /ex appears to be incorrect. * The chisquared test statistic and the pvalue using the option /ex is 6.147 and 0.0414. * If the test statistic is correct a comparison to a chisquared distribution of 1 degree of freedom should give a pvalue equal to 0.0132? * Possibly the correct pvalue should be calculated as 12*(0.0414) = 0.917? * EpiData Analysis (build 182) * The chisquared test statistic and the pvalue for a MH weighted odds ratio using the option /t is 6.15 and 0.046 * This pvalue using the option /t appears to be incorrect. * The pvalue using the option /ex appears to be incorrect. * The chisquared test statistic and pvalue using the option /ex is 6.147 and 0.0414. * If the test statistic is correct a comparison to a chisquared distribution of 1 degree of freedom should give a pvalue equal to 0.0132? * Possibly the correct pvalue should be calculated as 12*(0.0414) = 0.917? *********************************************** logclose 
This task depends upon