How do I prove my results are statistically significant or not?

[Soc]

Sorry in advance, I'm a complete statistics novice!
I have a three chemical samples which were prepared in exactly the same way, and when tested have given different results. How would I statistically compare these?
I was told by somebody to divide the average by the standard deviation to get the p-value..
I need to include statistical analysis in my report but I'm so confused as to how to do it.

Any help would be most appreciated,
Sarah

 

[noetsi]

Tested how (that is what result is generated)? Depending on what type of result is generated you can run different statistical test. Say for example you are measuring the ph of the three chemcials. In practice PH can be treated as an interval dependent variable. Than you have a dummy predictor (two of the three concentrations coded as two seperate dummy variables - the third concentration is analyzed through an intercept and is called the reference level. It has no dummy variable). If the p value for the dummies are signficant (below .05 normally) than that dummy is signficantly different than the reference concentration(the third concentration you did not code as a dummy directly) in terms of ph.

Dividing the average by the standard error may give you a t score (or similar statistic) to test a difference. It does not directly give you a p value (which has to be calculated given the score and an assumed distribution for the test). Regardless the simple answer is you run a statistical test

 

[Soc]

Thank you, I think I kind of get what you you're saying...
I've bound a drug to a protein, filtered it to leave the unbound drug and have tested the concentration of it through a HPLC machine. Which involves creating a calibration curve and then using the area of the peak and the equation of the calibration curve to find the concentration. I'm not sure if this can still be treated as an interval dependent variable?
Is there a software or something I can use to calculate the p value, because despite your explanation I 'm still not really sure how to do it.

 

[noetsi]

Generally if the distance between each level is the same, there are many distinct levels (12 or more is a reasonable rule of thumb although formally the answer is infinite) and the levels are ordered you can treat it as interval. I would think concentration, a percent normally, would certainly be considered interval. One possibility if you are unsure is to look at how past literature has treated concentrations.

No one calculates p values manually (well Dason might, but he is strange). Any decent statistical software will generate a p value. You have to decide what test you are running first. To me, assuming your dependent variable is interval which I suspect it is, then regression or ANOVA seems the most logical. ANOVA is commonly used in medical or biological research - not sure if that is what you are doing.

 

[Soc]

Sorry to be so dim but I really don't understand statistics. I tried to look at other literature but couldn't really work it out.

If I have three samples which were prepared in the same way and the results vary. They're measured in mcg/mL. I've downloaded Graph Pad Prism to try to do the statistics, but have only been able to calculate the mean, standard deviation and standard error of mean.

Is there a way to analyse the three results to say that they are reproducible and therefore good data?
Thanks

 

[noetsi]

I suggest either getting a commercial software such as SAS, SPSS etc (if you are at a university or medical facility the odds are someone has this). An alternative is learning R or trying excel's add on [although I am doubtful of excel personnally when it comes to stats]. Try linear regression with the results being the mcg/mi measure and your predictor being two dummies. One coded 1 if the first sample is present, a second coded 1 if the second sample is present and the third sample not measured directly as a dummy. It is the reference level. See if any of the statistical tests are signficant for the dummies.

Note in my comments above i assume the 3 samples are three different chemical samples. Not preparing the same sample 3 different times [which would be a different issue entirely].

 

[Dason]

I think noetsi is really jumping too far ahead here. From what I can tell you only have three values and they were generated in the exact same way but you get a different response. This is to be expected - there is almost always variation in data. However, I don't really see a question that can be tested statistically yet. If the research question is: "Is there a way to analyse the three results to say that they are reproducible and therefore good data" then you'll need to define what you mean by that.

 

[Soc]

My uni has been really unhelpful with statistical analysis, so I'm not sure if I should be using it in my dissertation!

My final experiment involved preparation of three samples of the same chemicals and doing quantitative tests on them. I understand there is variance in data so the quantitative tests are different.

I was unsure if I needed to state something about the variance and whether the results were reproducible and reliable?

 

[Dason]

Results being reproducible and reliable don't have a set statistical definition. You would need to define those a little better. If you aren't really sure what you mean by that other than a vague general sense of that that means then maybe you should ask your adviser for some direction with respect to that.

It might just be that you need to make a confidence interval for the true mean of the process.

 

[noetsi]

Per the last two comments I obviously misunderstood what was being asked. I thought there were three levels of a chemical [or different chemicals] that you were testing to see if they had an impact on some DV. Not that you were trying to understand why variation occured in the creation of a single chemical. I suggest following Dason's advice [always a good idea in any case].