Hi could someone help me to know what statistical test to use for comparing concentration of ellagic acid over time, I did three measurements of ellagic acid for each time (5hours, 10h, 15h, 20h) and I wanna see if there are differences in ellagic acid concentration through the time , I was thinking about using tukey test to see differences but I dont know if I´m right.
thanks
Hello,
The best way is to use MANOVA for repeated measures. In SPSS, the syntax command is something like that:
GLM time05 time10 time15 time20
/WSFACTOR=concentration 4.
Where time05 time10 time15 time20 are the four concentration measures for each subject and concentration is just a name given to the intra-individual factor (the 4 times).
If you are not comfortable with that multiple analysis, you can use bivariate analyses like paired samples T-tests but you will have to compute 6 analyses: 5h vs 10h, 5h vs 15h,..., 15h vs 20h.
In SPSS, the syntax command is something like that:
T-TEST PAIRS= time05 WITH time10 (PAIRED).
Best regards
CH
Hi Ricardo,
one way is to use Repeated measures anova (RM-ANOVA), which uses wide format of the data, i.e., one column for each of measurement (3 levels) of each time window (5-10-15-20 = four levels) = 12 columns per participant (row). Such RM-ANOVA would therefore have 2 within-subject factors (measurement and time window). Maybe you are not interested in the separate measurement within each time window as such, you could do averaging (then you would only use one factor = time window, which is the average of the 3 measurements). However, the averaging is not necessary.
Another way is to use Linear Mixed Effect Model, which is a bit more sophisticated approach (perhaps not that much needed in your case). These models usually require so-called "long format", i.e., each measurement/time window would be a separate row (as opposed to column). Therefore, each subject (marked in a separate column with a proper code) would have 12 rows, which are distinguished as belonging to the same subject by the code (i.e., "Sub1").
In some cases LMEM approach is superior, but as I see your design, RM-ANOVA is really enough! See an SPSS online tutorial on youtube: https://www.youtube.com/watch?v=6T6dvrwDe_U. Other software are usually also well-covered/documented there...
PS: i would not recommend using multiple t-tests if you are interested in a trend over time (i.e., generally increasing/decreasing values as a function of time) since t-test tests the difference between only two time windows, not the overall pattern of the data.
Best,
M.
Martin,
your proposal is very close of mine.
But if one is not comfortable with mutivariate analysis one can use the second option. Using all the combination of pairs of variables (5h, 10h, 15h and 20h) and computing all the t-test, you obtain exactly the same results. If your t is significant between 5h and 10h, you have an increase or decrease of concentration. If your t is not significant between 10h and 15h and not between 15h and 20h, you can conclude that the concentration increases or decreases during the five first hours and stay stable after. If all the t are significant, you concentration varies during all the period of time and you will have to analyze the direction of these changes: linear or not variation over the time.
Best
thanks everyone for your advice I'll try the test you said, tukey is ok for the experiment that I have?
thanks
- Badges
- Science topic
- Similar topics
- Cognitive Science
- Thinking