Very Confused

[Quisty]

Hi, I'm new to this forum, but I really need some help here.

I worked all summer on an experiment and I'm not sure what type of statistical test I need to perform on it. I've done linear regression analysis, but that isn't too reliable.

My variables are:
Distance (fixed, in km)
Rate (in percentages)

I'm looking at distance from a source and the rates of antimicrobial resistance to see if I can find a trend.
Does anyone know how to do this? I have three locations and six different antimicrobial types.

 

[JohnM]

Hi. Welcome to the forum.

Can you give us a bit more info?

-background/purpose of the study (1-2 sentences)
-what is antimicrobial resistance?
-were all 6 antimicrobial types in all 3 locations?

I think regression may be the method if you're trying to relate distance to resistance rate....but if there isn't a trend or relationship, then trying dozens of different statistical procedures won't help.

 

[Quisty]

More Info

Sorry for the delay, I've been very busy the past week.

Antimicrobial resistance is also called antibiotic resistance. It's when bacteria can't be killed by certain antibiotics.

I collected water samples in three cities along a river. For each city, I grew and collected bacterial colonies until I had 96 isolates per city. I then took these isolates and placed them on media containing one of six different antibiotics. If they grew, they were resistant to that antibiotic. I counted up the amount of growth on each antibiotic and calculated what percentage of the isolates were resistant.

I performed three trials. However, I could not afford to do genetic analysis, so I don't know if the bacteria was related or not. It's possible that I had the same species growing as different isolates.

So, basically, I have six antibiotics tested in each location. My null hypothesis would be that no bacteria grew, which would mean that the bacteria was not resistant.

It's complicated. I did do linear regression, but I really don't know if that is accurate enough.

Is that easier to understand? Do I need to clarify anything?

 

[JohnM]

Quisty,

If your null hypothesis is something as simple as "bacteria growth on a particular antibiotic = 0" then linear regression is not necessary.

All you need to show is whether the percentage of bacteria growth is statisitcally larger than 0.

Or, if you are comparing the effectiveness of each antibiotic, then you could compare the percentages to see if the difference between them is > 0.

Compute the average rate for each location or antibiotic or both, compute the confidence limit around it, and if the lower confidence limit is > 0, then it's significant.....same basic idea for the difference between two proportions.

Your study mechanics are becoming clear, but you stil need to develop a succinct and definite research question(s). Exactly what are you interested in learning from this exercise? Do you have a theory that you would like to prove or disprove?

Getting this written down will then lead to the appropriate statistical method....

JohnM