Hi,
I would like to compare the biochemical and cytological values like cell count, glucose levels, etc of test samples between two different tests, PCR and culture, for identification of Streptococcus pneumoniae. I would like to check if the values differed significantly among the samples positive by each test alone, and in combination. What statistical test(s) should I use for this? Is Fischer exact test with two-tailed p < 0.05 can be used? Kindly help me.
Thank you.
If you need to compare 2 numeric variables, you must to know if both variable have a normal distribution... If that hypothesis is ok, you can use a t-test to compare 2 samples or anova for 3 or more... Otherwise, if you reject normal distribution, you must use a mann Whitney-u test to compare 2 samples or kruskall wallis to compare 3 or more variables.
I don't understand you, the term "combination".
It indeed depends on the distributions. For 'cell count' this might be a Poisson distribution. You can get a first impression of the distributions, if you plot histograms of these for your variables/values. There are nonparametric alternatives for multi-way (combinations of independent variables) factorial designs. See attachments.
Dear MURTHY
The choice of statistical analysis depends on the type of your variables ( scale, ordinal, ..) and the distribution of your data ( normal distributed data or not ), that is will determine your choice for parametric or non-parametric tests.
Good Luck
Dear, Murthy
Even, if the distribution is not normal, we can transform the variables (log, square) and verify that the distribution has normalized and then used the student test which is robust. You can also use student test if your sample is large ( limit theorem) . The Student test compare the mean of two samples , in the case or the sample has a normal distribution.
best regards
- Badges
- Science topic
- Similar topics
- Chemistry
- Nanotechnology
- Nanoparticle Research