No acceptable conclusion can be drawn here based on only 20 studies in this case.

Refer this:

https://www.emerald.com/insight/content/doi/10.1108/BIJ-08-2017-0223/full/html

Hello,

You may consider using systematic meta analysis, which can be fortunately done by Stata 16. There are particular data you should collect from these studies.

You can read running meta analysis in Stata from here https://www.stata.com/new-in-stata/meta-analysis/

Regards,

Saed Sulub : Thank you for the insight. This is the answer I was looking for.

Greetings,

Before going into any statistical test, I would suggest making sure the conclusions are exactly based in the same criterion (either strain controlled or stress controlled fatigue tests... it is expected that stiffer mixtures - such as mixtures with recycled materials - would perform better in stress controlled tests). Many other complications may arise. For example, some conclusions on fatigue in the literature are based in non-homogeneous tests - where neither strain or stress are really controlled, but only the load applied. Also, some fatigue tests used in the literature may present more loading in one direction than in the other, then there is excessive creep during the test, what damages the specimens. In these tests, stiffer specimens tend to perform even better.

Try to link into the type of fatigue test and also in the stiffness of the material in the fatigue testing conditions. This may be of more insight than statistical tests.

Hope this help,

Lucas.

@@@@@@@

Hi Chamod Hettiarachchi ,

Thank you for this interesting question. First, I have to say that I agree with Lucas F. A. L. Babadopulos . The type of testing can bias your result. Another point of interest is whether those studies use similar types of mixtures (open, dense etc), bitumen type, bitumen content, and eventually have similar volumetric properties. All the abovementioned can strongly bias the result of your dependent variable (in your case mechanical properties).

If you manage to collect data that have similar characteristics and are tested in a similar way, and you still want to proceed with statistical analysis, you need to be careful about the type of statistical approach you want to apply. That means parametric or non-parametric statistical analysis. Parametric analysis is more robust, but 20 observations are maybe not sufficient to meet the prerequisite assumptions (normal distribution, no significant outliers etc). Since you dichotomize your independent variable (presence of RAP: yes and no) you can use a sample T-test (parametric) or a Mann-Whitney U-test (non-parametric) for each mechanical property you want to investigate. Another possibility is to split the presence of RAP into more than two levels (No RAP, low RAP and high RAP content). Then you will need to run an ANOVA (since you have more than two levels).

Various software exist out there to perform such analysis. Some examples are SPSS, JMP, R, STATA etc.

I hope you will find this somehow useful. Gool luck with your research!

Hi Alexandros Margaritis ,

Thank you for your valuable answer. I find your answer very useful and informative. I have read some of your papers and they are very innovative and interesting. Good luck with your research as well. Nice to e-meet you!!!

Chamod