Has anyone taken a look at the underlying data set, specifically for the compensation aspect? https://drive.google.com/file/d/1QOmVDpd8hcVYqqUXDXf68UMDWQZP0wQV/

Has anyone taken a look at the underlying data set, specifically for the compensation aspect? https://drive.google.com/file/d/1QOmVDpd8hcVYqqUXDXf68UMDWQZP0wQV/view https://drive.google.com/file/d/1QOmVDpd8hcVYqqUXDXf68UMDWQZ... Each record has a currency, a "CompTotal" (number of units of pay currency), a "CompFreq" (periodicity of pay; yearly/monthly/weekly) and a "ConvertedComp". ConvertedComp is CompTotal multiplied by the number of CompFreq periods in a year, converted to USD based on the specified currency and then clipped at the 98th percentile which seems to be $2M. It seems ConvertedComp is what analysts are intended to use. Of the United States-based, developer-by-profession responses who gave a salary (~13,500), about 14% specified their CompFreq as something other than annually. That subset has a mean ConvertedComp of >$1M. On the other hand, there are a grand total of 17 responses for ConvertedComp of >$1M for those who specified their CompFreq as annually. Color me suspicious, but I suspect a large fraction of the responses in that subset gave annual compensation rather compensation-per-period; undermining the calculation of ConvertedComp quite considerably. I don't know which fields the author actually used, but I have a lot of questions about the usefulness of this data.