I though that degrees of freedom for t test is the number of participants minus two parameters (means). However in the row for the equality of variance not assumed it could be for example 35, 598. Any explanation for that?
Thank you
I think it's due to the way that df need to be adjusted when equality of variances is not assumed: There's a good explanation at: http://stats.stackexchange.com/questions/116511/explanation-for-non-integer-degrees-of-freedom-in-t-test-with-unequal-variances
Which program do you use? Is it possible that the program automatically uses an adjusted test (Welch I guess), when the assumption of equal variances is not met? This would explain the decimals for the dfs.
Hello Rainer,
I was using SPSS, as on Stata the Levene's test does not appear directly.
Hope it helps
Witold
In SPSS you should see two rows, the first row with the "normal" t-test (including the Levene test) and the second row with the Welch test with adjusted df to correct for unequal variances. Does it look that way?
I attached an example how my output looks like (SPSS 23, german version).
Hello Rainer, I am not able to magnify the SPSS output you have attached
Witold
Same problem here, try right mouse button and "save target as". Then you should be able to save teh original .jpg file.
Hello Rainer,
Did that and what I can see is just "pixelosa". Perhaps you could use snipping tool to save the output.
Cheers
It has nothing to do with a snipping tool. It is cropped accordingly. It does not work on my machine when I use "save graphic as", but "save target as". With the former command the download ist only 2kb, because it saves the thumbnail. With the latter command it saves the original file, which should be around 68kb.
So...I have one more question about this topic...
So how do you report Welch's degrees of freedom? do you just round them?
I have looked for articles to get an example of how to report them, and so far I have seen they round them up, but I wonder if it's the right way.
Something like this?
(F2,426.67 =58.05, p= 0.001 or (F2,427 =58.05, p= 0.001 )
I think according to APA you have to report 2 decimals, and this is reasonable. Maybe rounding would be ok if you additionally report the eta-value, i.e. the deviation from homogeneity, so that every reader could caclulate the exact df-value.
- Badges
- Science topic
- Similar topics
- Biological Science
- Bioecology
- Systems Ecology
- Landscape Ecology
- Connectivity