Hello,
I have run an experiment adding ALA (four concentrations vs. control) to pancreatic cancer cells. To analyse the data, I'm using one-way ANOVA with Bonferroni correction. My explanation of the Bonferroni correction applied to my experiment in my thesis is as follows:
"If the one-way ANOVA result is statistically significant, post-hoc analysis in the form of multiple t-tests can be performed between all groups to determine the ALA concentrations at which the fluorescence is significantly higher than control. This requires 10 t-tests per time point (0.00 vs 0.25, 0.50, 0.75 & 1.00; 0.25 vs. 0.50, 0.75 & 1.00; 0.50 vs. 0.75 & 1.00; 0.75 vs. 1.00mM ALA). However, multiple t-tests increase the risk of a type I error (false rejection of the null hypothesis). If ɑ is set at 0.05, then a type I error is likely to occur every ~20 t-tests. The Bonferroni correction divides ɑ by the number of t-tests to compensate for this. In this experiment, the new ɑ to determine significance is 0.05/10 = 0.005."
My question is, when I look at the p values on the post-hoc analysis table that SPSS generates, should I be using 0.005 as my cut off for statistical significance as would be suggested by the Bonferroni correction? If the value says e.g. 0.01, then should I reject this as not significant?
Many thanks in advance.
Thanks Jos - I've just had a look at that tutorial. I didn't realise that the first column asterisks the significant values. Looking at my data, one of my rows that has a p value of 0.031 is asterisked. This suggests that the post-hoc analysis is saying that this is statistically significant. However, shouldn't the Bonferroni analysis should be using 0.005 as its significance level given that it should be dividing alpha by number of t tests (in my case, 0.05/10)?
Don't worry, I found the answer:
https://www.youtube.com/watch?v=pbWkwEz-XBY
Apparently SPSS maintains the p value in post-hoc analysis at 0.05 and instead multiplies the p value by the number of t tests being performed. In my case, it's multiplied the uncorrected p value by 10 (to correct it) and if it's still
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS