It is 30 years since I did Stats 101, so I need a bit of help before I revisit Stats through LinkedIn Learning.

I have attached an image on a spreadsheet whereby I have 10 leaders by 5 (A-E) variables, with the raw data and population proportion sample size. There is the REPORT which are the true numbers which is the system of record, and the SURVEY which is other individuals that reported their results accumulated. So we have two different subsets. I have been asked whether I need to scrap the survey if there is no statistical significance between the two, REPORT and SURVEY. If the leader 1 with variable A from the report is not statistically significant to leader 1 with variable A in the survey, and so on, I will scrap the report. I am looking at a confidence rate of 95%.

So the main questions that I have:
Should I use the population proportion sample size for the analysis?
There is no way to measure one variable over (Report) another one variable (Survey)?
Should I be looking at an Unequal Variance T-Test to test this, or some other test?

These variables are performance metrics for a leader.

If this isn't clear, please let me know.





TS Contributor
Why do you want to perform a statistical test here? There are huge discrepancies
between the data sources, especially regarding leaders 1 and 8. What difference
would it make to have the label "(not) statistically significant" attached to that table?

With kind regards

Thanks - If the Survey data is no where near a true representation (alpha <= 0.05) of the source of record Report data, I will be wasting my time managing the Survey.