doubt in understanding the code/procedure

[tinayong]

Hello All,

I have a doubt in the below program, can anyone help me to find why they have used 0.5 while calculating logauc differecne, and also how to interpre the results

data bioequiv;
input subject sequence $ AUC_T CMAX_T AUC_R CMAX_R;
logAUC_T=log(AUC_T);
logCMAX_T=log(CMAX_T);
logAUC_R=log(AUC_R);
logCMAX_R=log(CMAX_R);
cards;
01 TR  4321.7  360.11  4415.3  278.48
02 RT  2139.5  326.19  8846.2  605.45
03 RT  8295.5  578.66  7399.1  608.67
04 TR  5974.9  412.69  3071.2  196.19
05 TR  5121.9  423.98  7513.4  681.34
06 TR 18750.4 1359.86 17096.2 1191.91
07 RT  5609.7 1046.37  4498.8  392.45
08 RT  9040.9  733.80  7847.2  571.51
09 TR 21120.1 2403.67 22603.2 1449.63
10 RT  4865.3  669.17  8127.3  738.77
11 RT  7244.6  717.01  6249.4  364.80
12 TR  5543.7  409.05  4915.3  468.31
13 RT  9748.2  907.64 12218.5  884.73
14 RT 10257.4  995.68  6021.4  557.16
15 TR 10996.2 2434.96 15284.0 2509.14
16 TR  8031.0  476.31  2495.1  265.09
17 TR  6527.5  565.66 14736.4 1070.06
18 RT  6385.5  579.72  5118.7  483.04
19 RT  4077.1  297.07  2068.3  123.90
20 TR  7088.0 1334.96  6363.2 1064.23
21 RT 10689.3 1702.07 36108.5 2552.32
22 RT  4615.1  656.33  4551.9  403.34
23 TR  3792.8  441.33 10444.7 2057.38
24 TR  9429.5 1458.82 21239.7 1899.42
25 TR  4272.4  500.45  7247.8  945.48
26 RT  9442.7  905.35  3872.8  247.49
27 RT  7373.2  688.03  5522.3  449.56
28 TR  1883.5  139.61  3299.3  213.63
;
run;

*************************************************************************
* Construct the intra-subject differences (times one-half) within each  *
* sequence.  Then perform the statistical analysis.                     *
*************************************************************************;

data diff;
set bioequiv;
if sequence='TR' then logAUC_diff=0.5*(logAUC_T-logAUC_R);
if sequence='RT' then logAUC_diff=0.5*(logAUC_R-logAUC_T);
if sequence='TR' then logCMAX_diff=0.5*(logCMAX_T-logCMAX_R);
if sequence='RT' then logCMAX_diff=0.5*(logCMAX_R-logCMAX_T);
run;

proc ttest data=diff alpha=0.10;
class sequence;
var logAUC_diff logCMAX_diff;
title2 'Parametric Analysis';
run;

Thanks

[ballardw]

Who ever wrote the code provides exactly why they used 0.5.

*************************************************************************
* Construct the intra-subject differences (times one-half) within each  *
* sequence.  Then perform the statistical analysis.                     *
*************************************************************************;

Without any knowledge of the entire process or question asking for the original analysis I'm not sure I want to guess about the whole process or why half the difference was needed.

[SteveDenham]

Square root of the ratio leads to half the difference on the log scale.  No clue why they want this, though.

Steve Denham