I'm trying to find a test to correct for type 1 error due to multiple correlations (15 variables). Although Bonferroni is generally considered too conservative, I'm having trouble finding any suitable alternative suggestions. Further, SPSS does not seem to cater for correcting error during bivariate correlation analysis.
It would be great if anyone can advise on 1) which test/s to use and 2) how to implement these on SPSS (e.g. via syntax). Thank you!
I often use Holm's correction. Many argue that if you are considering Bonferroni's, you should use Holm's instead because it 'dominants' (statistical term, but means always at least as good) Bonferroni's. If you download R (https://cran.r-project.org/), then just type all the p values in from the correlation matrix (lower or upper triangle) as
ps
I often use Holm's correction. Many argue that if you are considering Bonferroni's, you should use Holm's instead because it 'dominants' (statistical term, but means always at least as good) Bonferroni's. If you download R (https://cran.r-project.org/), then just type all the p values in from the correlation matrix (lower or upper triangle) as
ps
Dear Daniel
Could you please inform us any suggestions or papers for using R, I have tried many times to deal with but I couldn't made it
That will be appreciated
thank you
Hi Rana,
Attached is Kami London's and my old intro. The R website lists a hundred of so books showing R. The intro one is available by going to Manuals on the CRAN page and it is the first one). The Journal of Statistical Software has about half of its papers on R, and the R Journal (https://journal.r-project.org/) is, well, all about R. For longer introductions, Field's Discovering Statistics series has an R version and this is a well liked introduction for both learning the package and statistics (note: I do like this, but most disclose I am the editor of the series it appears in). For specific questions (and hard questions), there are listservs for particular popular packages and Stackexchange has lots of R questions answer (so I would ask questions about R there, not on ResearchGate).
What level would you like?
Chapter A very brief introduction to R (ch. 1 of Wright & London, 2009)
I checked and there are a number of Excel based calculators available on the internet, if that would be more accessible for you than R.
Dear Daniel it's such an honor for me to got help from the editor himself :) thank you so much, hope you the best
Dear David thanks for responding, I would absolutely try this it might be much easier.
Thank you for the suggestions. If anyone else is struggling with this, I found Daniel's suggestion regarding the Holm's correction to be most appropriate. Thanks!
Thank you everyone. I am struggling with the same problem. Just I want to ask Jade if you could use R in this case easily or not? Could you please explain it more to me? As I understood, first I can get the P-values in the correlation table in SPSS, then ask R to give me the Holm adjusted p_values. is that right? the In fact, I have no prior experience working with R.
If you are doing the whole thing in R it is fairly simple. There is an R function in the package psych that does this. If you have installed psych (install.packages(psych)), and attached it (library(psych)), and let's say your variables are called v1,v2,v3,v4, then
corr.test(cbind(v1,v2,v3,v4))
will give you by default the Holm adjusted p values.
The SPSS software is slightly weak in this regard. I think the best solution is to use Minitab software and its modules.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS