I worked on 3 water bodies which are polluted by different sources and collect channa punctatus from each water body to check its health. I collected water and checked 7 heavy metals in these water bodies and their distribution in fish(n=26, collect from each site) gills, liver, kidney, muscle, integument (analysed in triplicates). Effect of heavy metal on blood profile (Hb, RBC, WBC, MCV, MCH, MCHC), lipid (total lipid, triglyceride, cholesterol, hdl, ldl) and protein (total protein, albumin, globulin) profile was also worked out (experiments conducted in duplicates).
I want to know which site is most polluted and also fish of which water body is most influenced. Is duncan will be suitable?
Dear Mehjabeen, with respect to your data you have 3 treatments and several variables (dependent variables) measured. For such data, you should check if your dependent variables are normally distributed, because in such studies there may exist dependent variables with non-normal data in their nature. When you separate normally distributed data you'd better run a MANOVA first, cos you've measured considerable amounts of DVs. Afterward, I recommend to use simultaneous multivariate comparisons.
For non-normal data in nature there exist transformations and other methods like Generalized Linear Models (GLM).
Actually the problem is that i have very little knowledge of statistics so you should have to guide me step wise step. Would you please tell me how can i check whether my dependent variables are normally distributed?
Dear Mehjabeen,
I had just written a long answer, but ResearchGate just swallowed it...sorry!
So here againg shorter:
You should start off by producing many, many graphs to see which patterns you find in your data. After all, the human eye is the best statistician.
After that, you may wish to calculate correlations between your heavy metal measurements and your response variables (blood cell counts etc.). This will give you a so-called correlation matrix. Check for values larger than 0.6 or smaller than -0.6 to filter out the strongest candidate correlations.
Finally, you may want to use simple linear models of the following form:
Response variable 1 = (Heavy metal1 + Heavy metal 2 + ... ) * Waterbody
Response variable 2 = (Heavy metal1 + Heavy metal 2 + ... ) * Waterbody
(...)
You should check the residuals of these models to see whether they show strange patterns.
Which software are you planning to use?
Best wishes
Christoph
Dear christoph, you pointed a good method. But the problem with this, is the multicollinearity between heavy metals as they have interaction effects. For this analysis I recommend Path analysis method. But as Mehjabeen said, he had little knowledge about statistics.
Mehjabeen, I recommend if you want to analyse your data in best way, you can have some statistician in charge to do all possible ways and find the best.
For normal test you can use Chi-square, Anderson-Darling, Kolmogrov-Smirnov, etc.
I would second the suggestion to find a collaborator with some statistical expertise. Unfortunately, there is no way to prescribe a "recipe" to follow for your analyses because there is just so much that would depend on the details in your data. Blindly following a prescribed recipe can lead to erroneous conclusions.
Well christoph scherber, would you please send your long answer to my mail id ([email protected]) or please give me your mail id so that i may contact and show you full data.
Dear B. Dar I too looking for DMRT. Please check is this the way of data feeding : If I consider Fe metal content in gill of fish of first, gill of 2nd site, gill of 3rd site and apply duncan. Similarly do it with another metal say Cu in gill of 1st site, gill of 2nd site, and 3rd site and apply duncan. Since i am studying 7 metals content in 5 organs of fish so do i carry out analysis in similar manner with other metals as well. This is done to check which which fish is more influenced. I am also interested to know that which metal is present in highest amount. So is this the way of feeding data on excel spread sheet: Fe content in gill, liver, kidney, muscle, skin of 1st site then Fe content in gill,liver,kidney,muscle,skin of 2nd site, Fe content in gill,liver,kidney,muscle,skin of 3rd site and apply duncan.
Please confirm this and tell me is this the way of feeding data in EXCEL SPREAD SHEET
Mehjabeen, be aware of the difference between your goal and your way of data analysis. I should note that finding which metal present most in fishes is nonsense, because each metal has its own threshold of damage.
If you want to totally compare 3 sites, you'd better use Multivariate ANOVA. this method consider all aspects simultaneously to compare water sites. If you want to find out which organ has most concentration of a metal, you should use two way (3 water sites * number of organs with organs as repeated measures and waters as categories) repeated measures.
If you want to consider each metal separately, you should use Analysis of Covariance. Because the concentration of one metal affects the others.
Hope it clear the cloud
EK
I think you want to compare some continuous variables in 3 groups. You can check normality assumption and based on it's results, ANOVA or Kruskal–Wallis test.
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis