Is it Ok to Analysis of the significantly / meaningfully of an inhibitory effect with anova-test or t-student?
We work on a compound which can decrease the cytotoxicity effect of another compound more than 20%. However, when we analysis about significantly, SPSS program indicates that it is not significant. It is not very rational. How can we limit the result between 20 and 25% or even between 5 and 15%. If a compound can decrease 10% of harmful effect, its effect is not significant and we should ignore it. Who determines acceptable range for being significant . Is there any standard for it?
The t-test may be used in most cases where you wish to compare any two experimental conditions. The ANOVA may be used in such a case, or when comparing more than two conditions. So, if you are comparing the cytotoxicity decrease with one compound versus another, then the t-test or ANOVA should work.
In your case, the t-test or ANOVA will compare the mean cytotoxicity decrease of each compound. The statistical test will also account for the data variation. Even if the percentage decrease seems to be large between the compounds, if there is a lot of variability in the data, then the decrease may fail to be significant.
Other data issues can cause results to turn up as not achieving statistical significance. I recommend first checking for outliers. You should also check to see if the different conditions have near equal variance. If one condition has a lot of variance relative to the other condition, this may well cause problems for a t-test, but especially an ANOVA.
Finally, I assume your question about significance relates to p-values? If so, there is little strict rule for what p-value is truly significant. In psychology, a p-value below 0.05 is typically what's needed, although a p-value between 0.05 and 0.10 might be reported as "marginally significant." Other fields have different conventions. I recommend checking any research methods texts for your specific field. It should mention what levels are significant.
I hope this helps!
It depends on how many factors your are analyzing. The t-test is a comparison between TWO means, used to compare group A with group B. If you have more than two factors, you would use an ANOVA. This would allow two factors/subgroups. It would be like comparing A, B, and C simultaneously rather than doing multiple t-tests (A to B, A to C, B to C).
If you already know that your compound decreases the cytotocicity, then why do you want to test anything?
If you instead want to estimate the size of the effect (reduction in cytotoxicity) and you have prior knowledge about a possible range of this effect, then you should consider a Bayesian method for estimation of this effect.
Significance is no objective concept. "Significant" is whatever the researcher can argue to be either
1) "a strange enough result under the null hypothesis so that it seems worth to be studied further" or
2) "a large enough effect to be considered (biologically, physiologically, economically, ...) relevant" or
3) "a result for which the null hypothesis can be rejected so that a long-run rate of false rejections won't exceed a given limit".
Argumentation 1 is from R.A.Fisher. He used p-values as a help to judge this "strangeness", but always in the context of the whole subject/reserch. Argumentation 2 is common sense; and when the variability of the result is considered, it it actually an extension of 1) for the case that the relevance of a finding can somehow be quantified at all. Argumentation 3 is a proposal from J.Neyman. This is the scientifically poorest justification because nothing is learned from the particular result and it is usually impossible to identify sensible error-rates (usually this rate is set to 5%, but if this is sensible in any particular case is more than questionable). However, this is typical approach in life sciences.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA