Graphical analysis for antibiotic sensitivity in Gram positive and negative bacteria

[garytse86]

I am doing an investigation on: comparing effects of antibiotics on Gram +ve and -ve bacteria.

I have used 8 antibiotics and I have measured the diameter of killing zones.

I am a bit unsure of how to carry out the analysis. I will be using Mann Whitney U Test, but what about graphical analysis? What type of graphs could I use?

Please help! Thanks in advance.

Gary

 

[garytse86]

anyone please?

 

[iansmith]

You could do a bar chart, having each antibiotic on the x-axis and the diameter on the y-axis. The bacteria would be group under each antibiotic.

You cannot use a scatter plot with the best line since you data do not come from the same sample. At least in micro, my profs were dictating that we use bar chart or scatter plot only.

Bar chart for data from different samples.
Scatter plot for data that come from the same sample such as growth curve.

 

[garytse86]

Alrite, thanks.

I was actually going to do a bar chart but I thought I might lose marks if I do so, so I asked to see if there are alternatives. Obviously I can't.

May be I can do error bars on the bar chart? Is that putting unceratainties on?

 

[iansmith]

If your data on the graph are an average of replicated and have enough replicates for each antibiotic for each bacteria, a standard error could be done and it would be better.

Put a standard error may support you data better. The smaller the better and this will influence you analysis. For example, base on your bar chart without error you conclude bacteria A is more resistance than bacteria B to a certain antibiotic, the error bar will often tell you if the difference has some statistical meaning. If you error bar of B is as high as the resistance level of A, than it is assume that A is no more resistant that B.

 

[garytse86]

So statistical tests like Mann Whitney does not take into account the random errors involved? I thought that is the whole point of a statistical significance of e.g. 5%, so I can be 95% confidence that the difference is not due to chance?

 

[iansmith]

The statistical test will support your observation and will give a more specific answer to your question because the test assesses if the degree of overlap between observed sample is less than the random expected value.

Following my example, if i said that A is equal to B with the graph, it because I am behind careful in my interpretation because I am using a qualititative method rather than an statistical test. The statistical test could said that the difference between and A and B is significant and that A is more resistant than B.

If you are doing a statistical test the graph may not be required but it sometime illustated and summarise your observations better than a statistical test.

 

[garytse86]

Ok. However I do have to include a graph in my project :( Although it does give a more clear picture generally, which is better than looking at a table with numbers. :)

Thanks for your help, really appreciate it :D

 

[iansmith]

As the expression goes, a picture is worth a 1000 words.

When you read paper, it is faster to look at figures rather than tables.