Why Quartiles calculed in stata are distint to quartiles in SPSS

[Sebastian Farias Hermosilla]

Hi.

I must calculate quartiles and must compare results with quartiles calculated in SPSS, but distribution always is different.

why ???


this quartiles are of a survey of housing and are quartiles from incomes



Thanks a lot

[Rich Goldstein]

as written, only someone knowledgeable about both Stata and SPSS could even begin to answer; I suggest you read the FAQ on how to ask questions, you then supply us with much more information and maybe someone can then help with the Stata part

[Clyde Schechter]

When, as is often the case in practice, the distribution function is flat over a range of values, quantiles may not be uniquely defined. Stata has one way of calculating them (see -help pctile- for details). It also provides the -altdef- option for another way. I don't know what method SPSS uses. You might try using the -altdef- option to see it the results match what you are expecting.

If not, you provide us little to go on for additional help. Showing us some of your data (particular a tabulation of values near one of the quantiles that you find problematic) along with the exact code you used and Stata's exact response (copied and pasted from the Results window into a code block--see FAQ for how to set one up if you don't already know) would be helpful.

[wbuchanan]

How does SPSS treat tied values? I recently had to try explaining to my IT shop how/why it was likely that Oracle would not yield identical results due to differences in the underlying algorithms. If you look at the help file for R's analog of pctile, you'd see that SAS also uses a distinctly different algorithm for estimating quantiles from SPSS. That said, I wouldn't be concerned unless the values were extremely different from one another. Oracle's functions, for example, are highly unstable depending on the normality of the distribution an concentration of ties; then again - and like I tried explaining - Oracle wasn't designed to estimate statistical functions.

[Richard Williams]

For those of us who do have SPSS buried somewhere on our machines, a replicable example, or at showing us the commands that were used in both SPSS and Stata, might give us more of a chance of answering you.