# What is the best statistical method for quantifying wastewater quality - (Apr/19/2012 )

There's no best method...it depends what you want to find out and also not, because you have different data types such as measured data (e.g. temperature, pH, electrical conductivity) and count data (if the bacteria and viruses are counted, not sure about that).

Anyway you can try to find correlations between different measurements such as DO and temperature or bacteria content and turbidity (though they sound quite commonplace for me).

Do you have any hypotheses for your data you want to prove? (e.g. a season or a certain treatment changes the values somehow or that interactions exist between different parameters....)

and as addendum:

For quality don't you have just to compare your data with threshold values that are given and if values are within the ranges (+/- some error range) it's okay?

What is the outcome of quality of waste water? is there any scale for the quality? a list of points? or quality test on fish survival?

If you can find a tool for quality of the waste water, a logistic regression analysis is a nice method.

hobglobin on Thu Apr 19 18:02:57 2012 said:

There's no best method...it depends what you want to find out and also not, because you have different data types such as measured data (e.g. temperature, pH, electrical conductivity) and count data (if the bacteria and viruses are counted, not sure about that).

Anyway you can try to find correlations between different measurements such as DO and temperature or bacteria content and turbidity (though they sound quite commonplace for me).

Do you have any hypotheses for your data you want to prove? (e.g. a season or a certain treatment changes the values somehow or that interactions exist between different parameters....)

and as addendum:

**For quality don't you have just to compare your data with threshold values that are given and if values are within the ranges (+/- some error range) it's okay?**

Unfortunately the treatment plant doesn't not have threshold standard they use (which I find unsual). I want to resist the temptation of using other international standard neither by the US, Europe nor WHO.

There is no working hypothesis for the data. Just a general hypothesis and the statistic aspect is just by the way. Any suggestive guide can you ? Thank you

newborn on Fri Apr 20 07:45:51 2012 said:

What is the outcome of quality of waste water? is there any scale for the quality? a list of points? or quality test on fish survival?

If you can find a tool for quality of the waste water, a logistic regression analysis is a nice method.

The research is just about to start, so no outcome yet.. The treatment plant doesn't have scale by which they access their effluent quality. We (I) are to develop one from the outcome of the research. Can please enligten more on the logistic regression method

Osu on Fri Apr 20 15:07:25 2012 said:

Unfortunately the treatment plant doesn't not have threshold standard they use (which I find unsual). I want to resist the temptation of using other international standard neither by the US, Europe nor WHO.

There is no working hypothesis for the data. Just a general hypothesis and the statistic aspect is just by the way. Any suggestive guide can you ? Thank you

That surprises me as I thought that there are general laws and regulations/standards every operator of such waste water treatment plant has to follow...From which country are you coming?

Anyway such standards can help you even if your plant ignores them: take them as the aim your plant should follow ideally and try to find out under which circumstances these standards are achieved and when not (i.e. they are within the limits or not)....

Statistics here could be that the deviations from standards are significant or not compared to a clean control sampled under same conditions (if you have this). But surely it's easier to take just given standards. Here modelling might be an option if it's not too difficult...it's then e.g. to predict future deviations from standards and how to avoid them by changing parameters (adding chemicals, increasing oxygen content, reducing waste water input, etc). This goes then to optimisation of the control of bioprocesses and chemical treatments...

But this depends what you want to find out and what the plant wants too finally and if there are such options...

Thank you all. Will come around again as soon as the numbers starts coming in...God bless

@hobglobin....Hello, compliments. First, thanks for your previous assistances and foresights. It got me started well. I was able to find a standard which guides the quality of effluents being produced. I have my data now and I have just started with the analysis but just using the basic descriptive statistics. Based on the previous discussion, you suggested using standards against the data; what statistical analysis is suitable for that? I am also considering correlating either the parameters as against another parameters or parameters as against microbial counts; what analysis will fit this? I will also appreciate any other suggestive analysis that can be done or carried out? Hope to hear from you soon.

A couple of statistical methods you might find useful to look into are autocorrelation and control charts.

since I don't know what exactly your research questions and hypotheses are and what you measured how often, it's difficult to suggest something. What DRT suggested is good to find out if a series of measurements has a correlation over time i.e. to find possible relations between measurements at different time points (autocorrelation).

To compare standards with your measurements you can most easily do a t-test (but the data have to fulfil some requirements such as normal distribution), and you need of course means and not single measurements (not sure if you have sufficient replications). There are some non-parametric alternatives too.

To compare over seasons for trends you might use the seasonal Kendall trend test.