Sonam Misra, I am not sure whether I understand the context of your research sufficiently to provide good advice, but from the information you've provided I think you could use all the data you have from each year group and perform an independent-samples t-test on those data.

@Sonam Misra Looking upon your query, I also agree with @Robert that independent samples t-test would be appropriate for your study. However, more information about your research questions/objectives and the type of data would help for better suggestions.

If it is survey data, i would guess that it is dependent data, iow the same subjects are asked both in 2019 and 2020, but with different response rates. If this is the case, I would recommend looking at those who have responded in both 19 and 20, and carry out a wilcoxon signed-rank test or paired t-test. Then, afterwards you should adress how those with missing data in one or both survey round differ from those who have survey repsonses in both waves.

If the two samples in 19 and 20 are independent I would recommend a Mann-Whitney test or independent t-test. Again, you should adress how non-responders differ from responders.

I am not sure that T-test should not be applied for ordinal data like the one you describe. But again, I am not sure whether it is a Likert scale with labels for each number on the scale or whether it is just a score of 1-10 on a continous scale.

Sonam Misra, from Jesper Fabricius's answer, it has become evident to me that you've not provided enough information for people to help you adequately.

Of some importance:

Did the 2019 and 2020 comprise different groups of people?

If the samples were drawn from the same group of people, are you able to match up the data from people who responded on both occasions?

What is the nature of your data? Did you simply ask one question concerning the training, or did you ask a number of questions and combine the responses from each participant to obtain a composite score for each of those participants?

Unless you provide information about those things (and maybe additional information - but I can't think of what else might be required at the moment), it is difficult to give you definitive help.

Hello Dear Sonam Misra

I support the answers to the above.( Robert Trevethan Jesper Fabricius Imran Anwar ) Because the independent sample t test, which is a parametric test, works on averages and this can be useful for you. However, you must meet the necessary prerequisites for your data set. However, I recommend that you also look at the paired samples t test. This test deals with the difference between pretest and posttest.

Best Regards...

Thank you everyone for taking out your valuable time and helping me out.

As suggested by everyone, here are some inputs-

Background - Training Style (curriculum) changed from 2019 to 2020 and I wanted to compare the effectiveness of the training for the 2 years. People were trained throughout the year. To complete a training it takes around 3 months.

Did the 2019 and 2020 comprise different groups of people?

Yes. It comprised of different group of people. A person who completed a training once will never repeat it again.

What is the nature of your data? Did you simply ask one question concerning the training, or did you ask a number of questions and combine the responses from each participant to obtain a composite score for each of those participants?

I asked multiple questions but there is one main question that captures how prepared they feel when they had to apply their training skills at work. I am considering that particular question to judge the effectiveness of training.

Any suggestions will be helpful?