Science Buddies: "Ask an Expert"

So I have finished the experiment and recorded all the important data. How should I do a Statistical Analysis towards the growth of bacteria in different agar petri dishes in seven days? What kind of data might be useful for that? Average, standard deviation..? Please help!

Hello EC314!

To do a statistical analysis, there are a couple of different methods.

I'd first suggest graphing the numbers of bacteria over time on Excel or a similar program. That should give you an idea of how the numbers changed over time.

Then, I'd suggest doing a regression; this is essentially a best-fit line or curve. In Excel, you can add a trendline; searching "add a trendline" should tell you how to do this if you haven't before. These can be linear, exponential, or polynomial, and each regression will also give you a correlation (R squared). The higher the correlation, the better the fit. If you try most of the options for a trendline and choose the one with the greatest correlation, you should be able to find the best fit. If it's linear, the slope will give you an idea of how fast the bacteria grow, and if it's exponential, the numbers in the formula will do the same.

You can also use the average number of bacteria over the seven-day experiment as a measure of how much the bacteria grew.

Thanks,
Valerie

Hey Valerie,

Thank you so much for the information! However, I still don't quite understand what correlation means. What do the different types of correlations represent? I do not have the bacteria count in each petri dish. However, I did record the fraction of area occupied by bacterial growth in each petri dish. Will that work as well? Thanks again for for your time.

EC

Hello EC,

Sorry for the late-ish response!

The fraction should work just as well.

Correlation is how well your data fits to a particular model. So if your data is linear, increasing at a regular rate, its correlation for a linear fit should be near 1. But if your data is linear and you try an exponential fit, your correlation will be more like 0.5-0.8, showing that the fit isn't as good.

Calculating correlation is a long process that involves summing up the deviations from the expected path as measured in an equally long way. If you need more details on that, I'd suggest Googling "statistics correlation coefficient" or something of the sort; I'm not too familiar with the process.

Hope this helped!