Protocol Online logo
Top : New Forum Archives (2009-): : Bioinformatics and Biostatistics

post hoc test LSD for multiple comarisions - how to analyze the results (Mar/04/2011 )

hi all

i have studied the preventive and curative effect of a drug on an animal model of disease. By applying LSD (spss software) for multiple comparison I found that both the treatments were sgnificantly different from test (animal model)but the p value was different. for example p value of some metabolite in preventive treatment group was <0.0001 highly significant were as p value of that metabolite in curative was <0.01. can i infer that preventive treatment was more effective than curative???

:)

-ujla80-

Hi there,

Two things, LSD is not a great post-hoc test. It is prone to false positives. At the very least you should do a protected LSD, or even better, a Tukey HSD test.

2) The answer is no on the comparison. The only way you can make the statement that the preventative was better than the curative is if you compare them directly to each other via the post-hoc. Based on the information you've given us you can only say that they are both better than the control group. You might be able to say there is a "trend" towards a greater effect in one versus the other, but unless the direct comparison is significant you really can't say anything.

MM

-Mighty Mouse-

hi MM

thanks a lot for help... :)





Mighty Mouse on Sat Mar 5 11:23:25 2011 said:


Hi there,

Two things, LSD is not a great post-hoc test. It is prone to false positives. At the very least you should do a protected LSD, or even better, a Tukey HSD test.

2) The answer is no on the comparison. The only way you can make the statement that the preventative was better than the curative is if you compare them directly to each other via the post-hoc. Based on the information you've given us you can only say that they are both better than the control group. You might be able to say there is a "trend" towards a greater effect in one versus the other, but unless the direct comparison is significant you really can't say anything.

MM

-ujla80-