Median vs 50th percentile comp

[nexus73]

I was curious what others are getting for compensation based on salary survey data. I was at a hospital for many years and every discussion regarding compensation landed at the median of the salary survey data, which was often around the 75th percentile. I recently interviewed at a hospital and they use the 50th percentile of survey data, which is notably lower than median.

I tried to talk with the psychiatry director about this and it didn't sound like a concept that had come up before. My understanding is that using the median is useful because it avoids extreme outliers from the data set. And in psychiatry especially, many doctors work part time or lower volume jobs which would pull down the average/mean/50th percentile number. Using 50th percentile instead of median seems like a good way for a hospital to skew salaries lower.

I was just wondering if I'm off base and using the 50th percentile is more common than median salary survey data.

---

Members don't see this ad.

 

[ObsequiousAplomb]

Your question is incredibly confusing. Why isn't the 50th percentile the same as the median on the data you're looking at? Median is defined as the 50th percentile.

 

[nexus73]

Your question is incredibly confusing. Why isn't the 50th percentile the same as the median on the data you're looking at? Median is defined as the 50th percentile.

yeah you're right, I was confusing 50th percentile with mean

 

[SmallBird]

I was curious what others are getting for compensation based on salary survey data. I was at a hospital for many years and every discussion regarding compensation landed at the median of the salary survey data, which was often around the 75th percentile. I recently interviewed at a hospital and they use the 50th percentile of survey data, which is notably lower than median.

I tried to talk with the psychiatry director about this and it didn't sound like a concept that had come up before. My understanding is that using the median is useful because it avoids extreme outliers from the data set. And in psychiatry especially, many doctors work part time or lower volume jobs which would pull down the average/mean/50th percentile number. Using 50th percentile instead of median seems like a good way for a hospital to skew salaries lower.

I was just wondering if I'm off base and using the 50th percentile is more common than median salary survey data.

Wait, so what is your question then?

 

[annoyedpsychiatrist]

I think you just confused the median and mean, if im not mistaken. The mean may be less accurate than the median if there are a lot huge variations that would skew up or down the average, so the middle value may give a more accurate representation in that sense.

 

[joooul]

It's a bad statistical practice to use any point estimate( Mean, Median, X percentile, Mode, QM, GM, etc..) to make inferences on any sample drawn from a population with an unknown distribution, such as salaries for psychiatrists. Ideally, you want to use a histogram/violin plot with superimposed individual points/box-plot with individual points, and then you can arrive at specific point estimates that can fairly describe the dataset. In medical salary survey data, they tend to use 25th, median, mean, 75th, and 99th percentile.

The main problem with any salary data presented to you by a hospital administrator is that the sampling process is unknown, and hence, even if you have access to "MOST ACCURATE SALARY SURVEY OF SPECIALITY X FOR Z YEAR BY COMPANY K" there remains a large chance that the data is not reflecting the reality because of the comprised sampling method (e.g., oversampling of large academic institutions, or large healthcare systems, undersampling of private practices, unknown $/FTE ) or admin just picking the survey data which fits them. There are multiple statistical techniques to estimate the true parameters of the population with outliers, but I don't think that the main customers of this data (hospital admins ) are interested in knowing the absolutely truthful distribution of salaries.

I highly doubt that even if you win the intellectual argument with the hospital administration and arrive at the correct numeric value for XXth percentile, you will necessarily get it. In the end, it all comes down to how much you value your work and how much they are willing to pay for it.