Deak Akila:

that is just the key problem! There is no way to do life sciences without a good background in mathematics. It is impossible to make people understand statistical concepts and statistical thinking who don't have a relatively sound math conception. To my experience, many courses and stats books - "for life science", "for dummies", "made easy" etc. -  make a bid deal in avoiding any math, what obviousely leads to a lot of misconceptions and confusion about the topic. Without some math, a student will not become literate in data analysis and interpretation (ei.e. not literate in statistics) and will ramain lacking some essential abilities of a scientist. As a result, students learn recipies they don't really understand and that they will often apply in wrong contexts. 

Get me right: I am not blaming the students. Most of them do their best. The educational system is bad. Math and stats are not really integrated into the life-science curricula. These are usually separate courses, and it is not taught as a part of the life-science topics (tha's similar with physics and chemistry for biologists, for instance: a biologist learns blood systems in a biology class without covering topics like fractal design and flow-dynamics, what will (or will not) be covered in some separate math and psysics class, and the student won't recognize that these things are actually important to understand the biology of blood systems). We lack real interdisciplinary approches in education.

And it is not really "higher math" I am talking about. To understand what probability is, some knowledge in measure theory is required, what in turn requires knowledge in set theory and calculus. I think, only measure theory is a bit beyond school math, the rest should actually be known from school. This is unfortunately not the case for most students (to my experience). And I have seen students (and graduates, post-docs and professors in life-science) who were unable to understand simple relationships like square-roots or logarithms. So there is really a lot of work needed to get students to have the background required to understand statistics (and empirical science). This leads me to the impression that most people seem to see anything beyond addition, subtraction, multiplication and division as "advanced maths".

You may have noticed by now that I can't recommend you a book that explains statistics without "advanced math". I also don't think that there is some easy way out. Understanding these things is hard work.

first, depends upon which type of data you have. there are different test for qualitative and quantitative data.

I agree with Dr Khanal, We should first know the nature of data before selecting a particular statistical test. I assume you have conducted an experimental study to see the effect of cyclotrisiloxane on MCF7 cells with ethanol as control. If your data is quantitative in nature. You can use paired t test for both experimental and control groups, using effects on cells at week zero and week 4.

If the effect you are evaluating is qualitative in nature, you should go for probabaly chi square test and other equivalents of T tests.

A statistical analysis is not the same as a statistical test. The former is very much about understanding the data, recognizing patterns, building sensible models, identifying effect and estimating their sizes.

A test is a bit a different subject. There are "significance tests" or "hypothesis tests", both of which follow different and incompatible philosophies. It looks as if you can do a "significance test", that is, if you have a sensible model calculate P(X|M) where X = "more extreme data" and M = "your model, resricted to the model-specific hypothesis to test".

All analysis starts with thinking about the data you are analyzing. What kind is it? What values can it take? How is it related to the biology? What can we reasonably assume about its characteristics or distribution?

The next step is to consider the experimental design: you have a time-series, you have several cell culture plates (was one and the same plate observed over time or were plates "sacrificed" for a measurement ?), did you prepare several different plates? from the same stock? did you use different stocks? What about the passages? etc etc etc. There is a hell lot to think about.

This will lead to some idea about a resonable statistical model that may be fitted to the observed data. This will give you estimates of effects that may be of interest for your research. In your case this *could* be something like growth-rate, or apoptosis-rate, or LD50 for some interesting parameter (like growth or apoptosis), or the time at which a maximum or minimum of the response is obtained, etc etc.

After having thaught about all this and after having an idea what is interesting for you, you might identify the important parameter in the model and calculate a "p-value" for its estimate given a sensible null hypothesis. This would be a significance test. However, it is usually miore interesting to provide a confidence interval for the estimate rather than a p-value.

Dear Meena

This information not enough to take decision about statistical analysis , what your question you want to test ? , do you take the period of exposure as a factor? , then are your data scale , ordinal etc, what the distribution of your data is ?

All these affect on your choice of statistical test .

Good Luck

in addition to above discussion, in statistics there are two types of tests -large sample and small sample tests  depending upon the size of sample which is drawn for a particular study. you may test a particular claim about the unknown parameter or you may test the difference between the sample statistic and unknown parameter on the basis of sample observations.

Dear

I am going to ask you what kind of data have you collected and what kind of results you need to validate: qualitative or quantitative? The type of test depends to final objectives of your experiment.

Thank

Marius

Marius,

the statistical test operates on a random variable. A random variable is a theoretical construct, a function turning data into numerical values (this is trivial when the data are already numerical values; however, the random variable may still consider a transformation) and having a probability distribution assigned to its domain.

Thus the question is not "what kind of data" you have but what properties the random variable has. Its domain may be discrete or continuous, and its probability distribution can have any shape that satisfies the conditions of probability theory.

Dear Dr. Jochen Wilhelm,

What book would you recommend for a life science research student to understand the concepts of statistics (parametric / non-parametric) from the beginning with a practical approach.? (Without advanced mathematics)

Your explanations in RG are easy to understand. Most of the time you give answers with examples and long explanations. I greatly appreciate that. 

I am glad if you can write a book with your vast experiences for life science students with less mathematical explanations.

Most of the books have written for the students who have studied applied and pure mathematics. Unfortunately, at present life science, students do not get a sound background in mathematics and this makes them difficult to follow statistics books available at present.

Deak Akila:

that is just the key problem! There is no way to do life sciences without a good background in mathematics. It is impossible to make people understand statistical concepts and statistical thinking who don't have a relatively sound math conception. To my experience, many courses and stats books - "for life science", "for dummies", "made easy" etc. -  make a bid deal in avoiding any math, what obviousely leads to a lot of misconceptions and confusion about the topic. Without some math, a student will not become literate in data analysis and interpretation (ei.e. not literate in statistics) and will ramain lacking some essential abilities of a scientist. As a result, students learn recipies they don't really understand and that they will often apply in wrong contexts. 

Get me right: I am not blaming the students. Most of them do their best. The educational system is bad. Math and stats are not really integrated into the life-science curricula. These are usually separate courses, and it is not taught as a part of the life-science topics (tha's similar with physics and chemistry for biologists, for instance: a biologist learns blood systems in a biology class without covering topics like fractal design and flow-dynamics, what will (or will not) be covered in some separate math and psysics class, and the student won't recognize that these things are actually important to understand the biology of blood systems). We lack real interdisciplinary approches in education.

And it is not really "higher math" I am talking about. To understand what probability is, some knowledge in measure theory is required, what in turn requires knowledge in set theory and calculus. I think, only measure theory is a bit beyond school math, the rest should actually be known from school. This is unfortunately not the case for most students (to my experience). And I have seen students (and graduates, post-docs and professors in life-science) who were unable to understand simple relationships like square-roots or logarithms. So there is really a lot of work needed to get students to have the background required to understand statistics (and empirical science). This leads me to the impression that most people seem to see anything beyond addition, subtraction, multiplication and division as "advanced maths".

You may have noticed by now that I can't recommend you a book that explains statistics without "advanced math". I also don't think that there is some easy way out. Understanding these things is hard work.