Dear Mahmoud

You can use regression test for finding correlation between PCR and BC.

As I understood, your response variable is dichotomous (yes/no). So a binomial model would be appropriate (generalized linear model of the binomial family, aka logistic regression or logit regression). The model can use the predictor "bacterial culture" as continuous (e.g. the bacterial concentration) or as categorical (e.g. yes/no or above/below some threshold) variable, whatever is available.

thanks for your concern but I hope you help me in this subject as I never use this test

have you studied the phylogeny of bacterial conserved gene sequences, if so then use MEGA 5.0 and try for TAJIMA test, it will give you an estimate about the correlation between genetic relatedness/ homogeneity of bacterial population. Use an outgroup member as a reference in this test.

if you have not sequenced the PCR products and you are studying through RFLP or any other method then use NTsys PC/NTedit software, there are many statistical tests for PCR based results.

As I understand you have two independent variables (PCR and bacterial culture) and want to assess whether there is an association with CMT. The most appropriate statistical method is chi-square or Fisher's exact test (if at least one cell of the contingency table is less than 5 observations). Also, you can conduct a test of agreement (kappa) which help to establish the correlation between the results of diagnostic tests. These methods are very easy to make and will give useful information.

Dear Mahmoud,

First, some points must be mentioned:

1. Results of CMT test are Positive or Negative. This test depends on the presence of white blood cells (WBC) in milk and in the case of subclinical mastitis the numbers of WBC are between 300000-900000 cells/ml milk. If you want obtain numbers, count WBC microscopically for each sample.

2. I do not know which PCR you used. Are the results some bands or anything else?

3. Are the results of the bacterial culture test numbers or anything else?

In case of normal data (numbers), use pearson correlation. In case of non-normal data (strong or weak bands, heavy or light growth) use spearman correlation.

Best regards,

Gaber El-Baradei

Dear Marcelo Signorini ·

good morning

I saw your comment that is right but I can not use the chi square test from before. so ask you if you have not mind to help me to perform this test with my results I will be thankful for you

Absolutely Mahmoud, I can help you but I need your data base to conduct the analysis. Best regards

MARCELO

please check your inbox I sent a message carrying the raw data

please accept my apologies for any troubles caused you