Statisticians prefer using R for many reasons. It is also free in the US. Software you're using is not as important as choosing the right method of statistical analysis: you would have to take into consideration a sample size, missing data etc. Just make sure it covers MANOVA if this is what you think you should implement.

They exist a lot of statistical software. I have good experiences with Statgraphics but you can try Minitab, SPSS, SAS, or some others

Try Design Expert. This is much better for such a purpose than a general statistics package such as statistica. http://www.statease.com/dx8trial.html

Design Expert is a great tool, but I am not sure if I understand your question: You need to find a software to run all the number, or analyze the number and make an interpreteation of them...

Design expert is a very good software.

try Cropstat, Genstat or SPSS

I work with SPSS, is a good alternative, but so is R or JUMP.

you've tried using contrasts

Observation:

It seems that in some cells in your design observations are missing : they can not occur by the very nature of your experiment.

In that case you should properly redefine the factors in your design.

Question:

Are you at all interested in between the factors in your design? If not, redefining your design as a design in which only one factor with 13 levels is present could be the answer. You could then use some one way analysis of variance, defining and testing the contrasts of interest. This can be done very well in SPSS.

I hope this helps,

Herman Adèr.

Thanks all for your valuable answers. Dear, Herman Ader, I am interested in the interactions between the factors that is why it become difficult to analyse such unbalanced factorial design experiment.

Dear Habte,

My remark is only relevant when cells are missing structurally. Even then, you can very well test interactions by properly choosing the contrasts you need An unbalanced design (without taking the structural missingness into account) can easily be handled by SPSS (ANOVA or GLM), For a design in which some factors have structurally missing cells you would need SPSS MANOVA to properly test the effects you need.

In R (previously called: S/Splus) you could you use lme and gls. R is free, of course and it is excellently documented (See also the reference below, but the online help functions are usually sufficient).

Myself, I often use MLwiN (multilevel analysis), it allows to define crossed designs, balanced or unbalanced, but, like SPSS, it is commercial software, although trial versions are available for a restricted number of days.

Herman

Reference

José C. Pinheiro and Douglas M. Bates (2000). Mixed-Effects Models in S and S-plus. Springer: New York.

Dear Habte,

Greetings!

Though I am unable to understand your query, because I am not well aware of these terminologies. I am using Graph Pad Prism. I can try to analyse your data using Graph Pad, if needed. You can contact me at [email protected] .

Regards,

Rajeev K Singla

SERB-Young Scientist

http://rajeevksingla.in

You can book an Econometrician from elearning.aneconomist.com and enjoy the learning of Econometrics, Statistics and Data Analysis. Let me do it for you. Email me on [email protected].

To few information on your problem: how many observations, repetitions in each cel (what does it mean unbalanced and incomplete - only structural missing observation or something else)?. In my opinion there is no need to write spectial syntax, it could be done using Statistica GLM dialog windows, choosing default general analysis options. When some interactions effects are not available because of lack of data, you can manually modify syntax of analysis, especially Design equation, according to syntax rules given in Statistica help.

Alternatively you cane use SAS 9.3.1 portable. Its syntax is very easy to digest. However, you will have to google the analysis type you desire. A general overview of the advanced (including incomplete designs) is present on this link. http://statistics.ats.ucla.edu/stat/sas/library/SASExpDes_os.htm#IBD

You can share you data on Researchgate as well or the structure of the data at least so that appropriate code could be attached for analysis.

Hope it helps.

Use SPSS or XLSTAT software which has free trial version downloadable. Its simple and self explanatory for the amateurs.

Last versions of Spss are more user-frienly than older ones. There is no more need to write syntax. You should just follow the sequential instructions in the proper window.

I would recomment to use R (http://www.r-project.org/) instead of commercial software.

STATA is another option in case you have access to it; you should be able to analyse your data in any of SPSS, STATA and SAS; even R.

To the best of my knowledge for a design in which some factors have missing i.e unbalanced incomplete factorial design, SPSS MANOVA can better test the variables of interest.

That should be quite possible, I can be of enormous help contact me on [email protected]. I will use either STATA, SPSS E-Views depending on the nature of your data

And bettern help than all is freely available at spoonful scale on UTCL website. Check it here http://www.ats.ucla.edu/stat/

If you want to use R, there is an excellent package that handles time-series panel data and unbalanced designs.

R has a wide scope of statistical evaluation, the software allows you to design and code your own test which in essence allows you to create your own tailor made statistical package for your project. R also allows models to be created and linked with other R features so that is a plus. On first use R does look difficult to use but after you know some key lines of code it is very easy to build upon this crating more advanced codes. In terms of what is better SPSS or R it depends on your field I believe SPSS is still the top software in the social sciences and is the best tool to teach statistics to all undergraduate students (physical & social science). But I do believe that at doctoral level researchers and above in the physical sciences and engineering should be using R or one of its equivalents

MANOVA in MicrOsiris performs an exact solution with either equal or unequal numbers of observations in the cells. The method applies to any design with proportionate or disproportionate cell frequencies and with or without covariates.

remember that the sample size is important in the multivariate analysis, especially in the MANOVA

Spss- MANOVA-R-spoonful scale on UTCL website-STATA all these softer could be useful for use .

I would suggest you look at the GenStat Software to solve your problem. All the above software applications are excellent but Genstat is designed for solving problems like yours.

Statisticians prefer using R for many reasons. It is also free in the US. Software you're using is not as important as choosing the right method of statistical analysis: you would have to take into consideration a sample size, missing data etc. Just make sure it covers MANOVA if this is what you think you should implement.

As far as I can tell, your issues have more to do with the realities you are trying to model than with the software you're using. I expect you to be able to obtain results when you leave out all of the interaction terms, and that you'll have a healthy suspicion of the results. It somewhat defeats the purpose of running an experiment that includes all of your treatment combinations but have you considered testing your interactions using only subsets of the data where the interactions are possible? If these tests of the subsets do not indicate significant interactions then you have a reason to publish tests of the main effects. If you use this analytical approach, please describe it in your publications. I hope my comments are helpful to you but I, like your other respondents, have failed to give you any Statistica syntax. My software of choice is R though I have a lot more experience using SAS because of requirements from my previous employers. If I need to provide more detail or define terms in my proposal, please let me know.