0
$\begingroup$

My research is question is: does the temperature at which a certain species of bacteria is incubated affect its growth?

I have two agar plates, each incubated at different temperatures. I counted the number of colonies in each one and got counts of 14000 and 1000. Since count data is discrete, I know I should use either a GLM or a chi square test. However, I am unsure which one I should use.

Could someone please help me choose the most appropriate test? I would really appreciate any advice.

Edit: my dataset looks like this:

               colony count at each temperature
Experiment no.   20 degrees     30 degrees   
1                 14000           1000
2                 12100           800
3                 11500           1200
Improve this question
$\endgroup$
8
  • 6
    $\begingroup$ Usually bacterial counts aren't treated as counts of individual entities. If you were sampling from environmental waters, they might be treated like a concentration (cfu / 100ml). In any case, if you have just two measurements, there is probably no value in conducting a statistical test. $\endgroup$
    – Sal Mangiafico
    Commented Jan 26, 2020 at 9:35
  • $\begingroup$ @SalMangiafico As mentioned in my edit, I have similar counts for two more sets of plates. $\endgroup$
    – user271904
    Commented Jan 26, 2020 at 12:09
  • 4
    $\begingroup$ What is your research question ? $\endgroup$
    – Robert Long
    Commented Jan 26, 2020 at 12:34
  • $\begingroup$ Even with N = 4, nothing is likely to be very effective, statistically. But please answer @RobertLong's question about your RQ (and put that in your question). $\endgroup$
    – Peter Flom
    Commented Jan 26, 2020 at 13:23
  • 3
    $\begingroup$ @PeterFlom-ReinstateMonica I think N might be 3 ! Anyway, working on the assumption that the RQ is along the lines of "are the counts different on the two plates ?" the question states that the other plates have similar counts, so with around 1,000 vs 14,000 this should be a case of common sense prevailing over the need for any statistical test. Of course they are different :) $\endgroup$
    – Robert Long
    Commented Jan 26, 2020 at 13:34

1 Answer 1

Reset to default
1
$\begingroup$

Since your research question involves a dependent variable (number of colonies) and an independent variable (temperature) you want regression. But you a) Should include all your plates in one regression and b) Will almost certainly need more data to get useful statistical tests, unless the relationship is very strong and clear.

I am not sure what "sets of plates" means. That is, are these sets somehow disginguishable from one another (different species or something?) If so, that will have to be dealt with as well.

It's true that OLS regression demands a continuous dependent variable and you have one that is counts. For count data, you can use Poisson or negative binomial regression. But, if all the counts are this high (thousands) then it probably won't make much difference.

Improve this answer
$\endgroup$
4
  • 1
    $\begingroup$ (+1) nice, Peter ! $\endgroup$
    – Robert Long
    Commented Jan 26, 2020 at 15:03
  • $\begingroup$ Thank you so much for the detailed answer. When I wrote sets of plates I should have been more clear but what I meant is I repeated the same experiment with the same species two more times. Could I please ask you to explain why a chi square test would be unsuitable? Is it because there is only one independent variable? $\endgroup$
    – user271904
    Commented Jan 26, 2020 at 15:16
  • $\begingroup$ @ulfelder No we only used two temperatures so I think I will do a chi square test. $\endgroup$
    – user271904
    Commented Jan 26, 2020 at 16:01
  • 2
    $\begingroup$ @Anya, You have a very good answer here. Your best bet is to set the independent variable as either a continuous variable or a 2 factor categorical variable and your dependent variable a continuous variable. Therefore either regression, or t-test is a better bet. In no case is a Chi-square test a good choice for this problem. See rcompanion.org/handbook, the section on choosing a statistical test. $\endgroup$
    – Dave2e
    Commented Jan 27, 2020 at 16:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.