In my opinion it has to be clear, "how do you measure" what is called "learning", but assuming that the measurement is done by placing scores on the answers of the surveys, we first compare the arithmetic means, of the pre and post of the questionnaires to see if there are significant differences in learning by the application of each strategy separately and then we compare the arithmetic means of learning between the two groups.

You can use a two-group repeated measures Analysis of Variance. Pre/post is the repeated measure, and live/recorded is a between-subjects variable. Check you distributions first, although RM-ANOVA is fairly robust. Not all statistical software packages include RM-ANOVA in their basic packages, but if yours doesn't, perhaps there is an R package available. RM-ANOVA will give you a lot of information if you are able to meet its requirements: whether the instruction had an effect (the repeated measure), whether the two groups differed overall, and--most importantly--whether there was a significant interaction term (e.g., the two groups were the same on the pre-test, but different on the post-test).

This is is tricky question. If you are collecting data feedback data in rating scale then you can look for the process here:

Article ELEI - METRIC FOR MEASURING ELEARNING EFFECTIVENESS

Also, you can conduct proportion test; Fisher test to find out if positive response percentages for the questions are greater in second case i.e. the post scenario.

For more discussion, you can reach me at:

[email protected]

For a little more detail, the surveys were the same pre and post for each training whether it was presented live or recorded. There is a total of six different training webinars I will compare. Each had their own pre/post-survey. I want to compare each one individually and then combine all data for an overall difference. The answers were (strongly disagree, disagree, neutral, agree, and strongly agree). I am also having a hard time figuring out how to determine the needed sample size for the recorded webinars to achieve reliable results. The live webinars averaged 250-300 participants.

Then you should use a RM-MANOVA. I was going to run a power analysis for you, but my Power & Precision software isn't loading properly & I will need to contact tech support. Perhaps you have some other power analysis software?

The best method will be the one you have developed for your specific task. This will make it possible to take into account the most important aspects of the study.

The best method that is appropriate in teaching statistics is to use software package after all the rudiments must have been explained to the students. Use of software is more accurate in teaching statistics

I am also having a hard time figuring out how to determine the needed sample size for the recorded webinars to achieve reliable results. The live webinars averaged 250-300 participants.

Quick cheat: use a sample calculator.

https://www.surveysystem.com/sscalc.htm or https://www.calculator.net/sample-size-calculator.html

also, for your first question.. another quick cheat for attitudinal data only: We use Qualtrics as it does the calculations for us and provides an overall report to include graphics. For our informal educators, it saves us a lot of time and effort and gives our stakeholders instant information. https://www.qualtrics.com/education/

Hope this helps.

Let's translate the problem into the language of mathematical statistics. There are two samples: before and after training. Let us form two functions of distribution from them: of the form: strongly disagree - the number of responses, disagree - the number of responses, neutral - the number of responses, agree - the number of responses, and strongly agree - the number of responses. After that, a well-known problem in mathematical statistics: compare two distribution functions.

The easiest way to compare is frequency analysis. Let's calculate the probabilities of each type of answer "before" and "after". We get tables strongly disagree - the probability of response = number of responses / (total number of respondents * 5) and so on. After that, you need to use the chi-square test. You can use any specialized package for processing data such as SPSS.

There is a great tendency to use social networks to research statistics.

There is no 100% credibility, but in "Net Times" we are subject to this risk.

I also use field research and IBGE.