Which statistical test to use for my data analysis

Discussions concerned with knowledge of measurement, properties, and relations quantities, theoretical or applied.

Which statistical test to use for my data analysis

Postby Biosapien on October 15th, 2016, 3:04 pm 

Hi everyone,

Could someone help me to know which statistics should i use for my data analysis.

I did my project on "personal exposure to PM2.5 among different villages". Sample size is 400 from 30 different villages. Among these 400, 40 subjects have 6 time PM measurement and the rest 360 subjects have 2 time PM measurement. The sample size among each village is different. I am curious to know which statistical test should be used to report the personal PM2.5 exposure among different villages. I also want to report variation of personal PM2.5 exposure among the 40 individuals with 6 time measurement based on month wise and season wise (pre and post monsoon). Thank you.
Biosapien
Member
 
Posts: 116
Joined: 11 Mar 2015


Re: Which statistical test to use for my data analysis

Postby BioWizard on October 15th, 2016, 4:10 pm 

Hi Biosapien,

You don't need a statistical TEST to merely REPORT data. You need a statistical TEST to >TEST< a hypothesis (hence, statistical TEST). Do you have a hypothesis?
User avatar
BioWizard
Forum Administrator
 
Posts: 12691
Joined: 24 Mar 2005
Location: United States
Blog: View Blog (3)
Natural ChemE liked this post


Re: Which statistical test to use for my data analysis

Postby Biosapien on October 16th, 2016, 12:44 am 

my null hypothesis is age does not influence the exposure to particulate matter.
Biosapien
Member
 
Posts: 116
Joined: 11 Mar 2015


Re: Which statistical test to use for my data analysis

Postby BioWizard on October 16th, 2016, 10:06 am 

Biosapien,

I'm shaking my head over here. You wanted to perform a statistical test to test a hypothesis concerning age, and yet you didn't even mention the word age in your first post. This, to me, reveals an alarming amount of confusion about how hypothesis testing works, and the role statistics is meant to play in the process. That said...

I think you need to restate your question, beginning with the hypothesis (which you should've synthesized before even beginning to collect your data). Then you need to build on that to explain why you collected the data you chose to collect, in the manner you chose to collect it (what EXACTLY did you measure? How? Who took the measurements?). After that, you should specify which - of the many parameters you mentioned (individual characteristics, village, location, etc) you think may be relevant (or not) to your hypothesis, and what parameters you'd like to test for significant influence on your outcome measure. Is it just age? Does village matter? Does person taking the measurement matter?
User avatar
BioWizard
Forum Administrator
 
Posts: 12691
Joined: 24 Mar 2005
Location: United States
Blog: View Blog (3)


Re: Which statistical test to use for my data analysis

Postby doogles on October 16th, 2016, 5:03 pm 

Biosapien, I notice that it was a "project" you were doing, not necessarily serious research, and that the title was simply '"personal exposure to PM2.5 among different villages". You may just need a Table of village averages if that really IS the case. As Biowizard has said, we need much more information about the project - as to whether total populations for each village, sex of individuals, smoking habits, prevalence of sources of dust (dust storms, roads, vehicle exhausts, cattle herds in dry conditions etc) if you need to identify factors associated with PM2.5 measurements.

The way you have worded the query suggests that biological samples were taken from individual subjects and measured. It arouses my curiosity as to how a figure was obtained for individual exposure to particulate matter less than 2.5 micron in size. Are you able to describe, or give a link to, that test. I'm personally quite interested.
doogles
Member
 
Posts: 861
Joined: 11 Apr 2009


Re: Which statistical test to use for my data analysis

Postby Biosapien on October 19th, 2016, 6:35 am 

Hi Biowiz,

I am interested to know which age group of people are more vulnerable to PM2.5 exposure. What i have is the demographic details and their time activity recall data. I am interested to see which age group of subject are highly exposed and for what reason (type of occupation, fuel used for cooking, kitchen type and so on). So my null hypothesis is, age is not an influencing factor for high PM2.5 Exposure and the alternative hypothesis is vice versa.
Biosapien
Member
 
Posts: 116
Joined: 11 Mar 2015


Re: Which statistical test to use for my data analysis

Postby Biosapien on October 24th, 2016, 12:09 am 

Hi Doogles,

I was collected the particulate matter through standard Gravimeteric Method using air sampler which can sample 5L of air per minute. While sampling the particles from the air will be collected and deposited on the special filter paper (PTFE), which will be pre and post weighted before and after sampling to estimate the PM dust weight. Later this dust weight will be used to calculate the PM concentration along with other parameters and formula.
Biosapien
Member
 
Posts: 116
Joined: 11 Mar 2015


Re: Which statistical test to use for my data analysis

Postby doogles on October 26th, 2016, 3:13 am 

Biosapiens. Thank you for that extra information.

It is in my nature to do my best to help anybody in any way I can. I am NOT a statistician, but I did have a period in my life when I understood and practiced statistics in helping scores of postgraduate students many times daily. But I’m having trouble in getting a full mental picture of the experiment you have performed. I suspect , like Biowizard, that your experiment has been poorly conceived and executed, and that you are not going to be able to achieve any meaningful results – THIS TIME. But don’t let this discourage you.

We (I include myself) all still learn from our mistakes, so maybe some good will come out of it. You appear to be willing to ‘have a go’ at doing something. I applaud you for that and urge you to keep doing it

Obviously now, you collected multiple AIR samples in a specified and consistent manner and assayed them quantitatively for the amount of Particulate Matter <2.5 micron in size. You took 400 samples altogether. There were 30 villages, but I have no mental image as to whether they were African thatched huts or English Hamlets, or whether you took samples next to saw mills, old folks homes, or kindergartens. Every single detail is important to developing meaningful interpretation of data.

Please don’t regard these comments as derogatory. They are made in the sense of explaining my own difficulty in being of assistance to you.

The first and most important statistical step that you should have made is to establish a ‘Coefficiient of Variation’ (CV), which is a test for the consistency and therefore reliability of the results you get from any analytical process for the test itself.

ALL RESULTS BASED ON AN ANALYTICAL TEST ARE USELESS IF THE CV for any given analysis is greater than the range of results in the collected data. All of the postgrads I had to co-supervise were trained to collect, and prepare their own samples in a specified and consistent manner and do their own laboratory analyses with every available machine. The first thing they had to do was to establish this Coefficient of Variation for 10 consecutive analyses of a single sample.

You can still do this by taking 10 consecutive samples of air with the same equipment in the same place as rapidly as possible.

Another factor in any analyses, all of which are based on physics or chemical principles, is that the machines tend to ‘drift’, that is, they tend to change in their readings – for multiple reasons. So, for reliable readings, they need to be checked and re-calibrated periodically with some sort of standard sample of known concentration. I cannot personally imagine how you could do such a thing with an air sample, but I would be surprised if such a method is not available in some form in research circles.

I’ll have to leave it there unless I can get a better description of the nature of the villages, standards of selection for sampling sites, demography etc and wish you all the best in your endeavours.
doogles
Member
 
Posts: 861
Joined: 11 Apr 2009



Return to Mathematics

Who is online

Users browsing this forum: Google [Bot] and 7 guests