I am looking for statistical analysis program that do analysis for 2 factor at 3 level full factorial design, and also do optimization.
We tried ANOVA univariate analysis for our experiment but its results of all the tests were insignificant.
Have you considered the fact that they may be truly insignificant.
You could try ezANOVA, its a freedownload statistical program.I use it for my data with brilliant results.
Regarding previous results analysis (ANOVA univariate analysis) we are hesitant because visually there are big differences between tested groups.
We will try using ezANOVA to confirm the analysis.
Thank you very much for your help.
You won't get different results from different programs, you get different results from different tests. You'll get the same results from univariate ANOVA regardless of what program you run it in.
As Dr. Onaolapo suggested, the fact that your results didn't come out significant probably just means there aren't significant differences. Fishing for other tests to use to get p
One reason why you might have insignificant results is that you don't have enough power as ANOVA is very sensitive to sample size. Did you do a power analysis (preferably prior to collecting the data!)? Now adding participants to get a significant result is a highly questionable approach but if your sample size is simply too small then you won't get a significant effect with ANOVA. You have also not mentioned if it is a within subjects design. If it is, you likely have sphericity problems and will find MANOVA more powerful.
These days my preferred approach is to use JASP and calculate the Bayes Factors. This will indicate the evidence for the Null and Alternative Hypotheses (it's a ratio of these two alternatives). JASP is free but a work in progress it is however, very easy to use. See: https://jasp-stats.org
In the study we had two independent factors affecting the parameter measured with 2-3 divisions within each factor, using univariate analysis we compared results of divisions within each independent factor alone, so we thought to make sure of the results of the statistic by doing analysis for the 2 factor at the same time (as mentioned before we hesitated to take results because there was big difference between groups visually) using soft ware that do analysis for 2 factor at 3 level.
Finally we tried ezANOVA which helped us do analysis for 2 factor at 3 level, and there was no significant difference, and we concluded that the difference that we considered big without statistic between experiment groups, was actually not statistically significant.
Finally thank you all for your help, we really appreciate it.
Dear Leena
Do you take in account that it may be the interaction between the two factors that may hide the true effect of each one?
Dear Leena,
If your 2 factors are qualitative, you must use ANOVA. If not, you can use regression.
Having performed a full factorial design, you can estimate main effects and also the interaction effect. In ANOVA, it means that you can test the effects of each couple and in regression the quadratic effects and the interaction effect (x1*x2).
That is incorrect. Regression is possible with qualitative factors (indeed, ANOVA is just a sub-type of the general linear model from which regression comes).
Leena Abdulaziz
See these packages from R
1- FrF2
2- BHH2
2- BsMD
3- AlgDesign
To answer the original question,
If you are going to do analyses like two-way anova, you'll need to get a good statistical package, invest the time to learn how to use the package, and learn about the assumptions and uses of the statistical tests you want to use.
Since you tagged biostatistics, you might be interested in SAS or R. SAS is relatively easy to use, but is expensive. R is a little more difficult to learn, but is entirely free. The deciding factor might be which is used in your university department.
For a free introductory text on analysis of experiments, using either SAS or R, I might recommend the Handbook of Biological Statistics ( http://www.biostathandbook.com/ ). It’s light on theory, but does a pretty good job covering the uses and assumptions of each analysis.
It’s also a good idea to get your hands on a good statistics reference book. Zar, Sokal and Rohlf, and Agresti, listed at the following link are good. http://rcompanion.org/rcompanion/a_10.html There are also some free statistics textbooks listed here. http://rcompanion.org/handbook/A_04.html
- Badges
- Science topic