In my study i have 2 groups of patient (diagnosed with stage 1 and 2 at the beginning of treatment), where dependent variable is "lesion volume", which was measured before and after treatment. Treatment time for each patient is not the same. I want to compare the response to treatment between 2 groups.
What is the best statistical test to use to compare compare the response of lesion volume to treatment? Is Mixed ANOVA test a possible test for my study?
You should test for normality before. In case the distribution is normal, you should use paired t-Test because you want to test each patient after treatment with himself before treatment. The pairing parameter is the identity of the patient. If the distribution is not normal your could try some transformation to get normality or choose a nonparametric test.
Dear Antonio Ruiz
Paired T-test is used to test compares the means of two measurements taken from the same group. But i have 2 groups, each group has 2 measurements (pre&post treatment) and i want to compare the change of lesion volume between 2 groups
At first I try to calculate the difference between 2 measurements (by simple formula: "Measurement_at_T1 - Measurement_at_T0) and label it "change of volume", making it into a new variable, then use Independent T-test to compare between 2 groups. However, some people think this way will not give result of change precisely.
Then it should be a t-Test between the two groups. Test the normality of the change in lesion volume. It's noise is the sum of the noise of pre&post volumes. Maybe the relative change (pre_volume - post_volume)/pre_volume would behave better.
It depends on but you are advised to check with www.Stats4Edu.com
The first step is to test for normality of the data (distributions) If normally distributed then, paired t test should be used.if the variables are not normally distributed then use a non-parametric test -Wilcoxon test also known as Wilcoxon signed-rank test.
Check on the normality of your data. If the data exhibits that the parent population is normal, then you can use paired t-test. Otherwise consider it nonparametric counterpart like ordinary sign or sign rank test.
Kemei Kimutai
My purpose is not to compare the result of the same group before and after treatment, but I want to compare 2 separated group in term of change between T0 (at base line) and T1 (after treatment)
Thank you mr Antonio Ruiz for your previous answer, below i have some more questions, please help me.
1. "t-Test between the two groups" is independent sample t-test you mean right?
2. "Test the normality of the change in lesion volume": here, should I test the normality of "the change in volume" based on 2 groups or test the whole data set?
3. "It's noise is the sum of the noise of pre&post volumes.": I don't understand what you mean with the word "noise"
4. "Maybe the relative change would behave better.": Yes sir, I also try this formula (pre_volume - post_volume)x100/pre_volume to calculate the change of volume in percentage. What I wonder here is, with this formula, we can get mixed values (negative value in case of volume decrease & positive value in case of volume increase). So, does this mixed (negative and positive) values affect the result of t-test, sir?
The attached file show my result of independent sample t-test with the above formula, I assumed that the mixed value data does not affect the test and my reading for value of -23.14 is "the volume decrease by 23,14% in stage 1", is this correct?
Thank you so much.
Dear Huy Quang Nguyen
1) t-Test between the two groups is independent sample t-test
2) each group should be normal distributed
3) By noise I mean the measurement error of the lesion volume.
I don't understand what Stage 0-1 and Stage 2-3 mean. Are they the 2 groups? In this case, for Stage 0-1 the volume decreased and for Stage 2-3 it increased. In both cases, the measurement errors are large but there is a significative difference p
3. I stiI have 2 basic groups in this study. Stage 0 and stage 1 are grouped together, stage 2 and 3 are grouped together, not a big deal
4. Now i understand the "noise" term that you mention is standard deviation, if i am not wrong.
So here I did paired T-test separately for stage 0-1 and stage 2-3 (the result is in attached image):
- there is significant difference for Stage 0-1 lesion volume before and after treatment (p=.016)
- while there is no significant difference for Stage 2-3 lesion volume before and after treatment (p=.153)
Until here, I can conclude that there is real change in pre- and post-treatment volume for group "Stage 0-1", and no real change in volume for group "stage 2-3".
Does the result above affect the result of the independent sample T-test
f you have good reasons to asume that the volume should reduce, then you can divide the p value corresponding to a 2-tailed distribution by 2 and take p=0.008, 1-tiled, (for Stage 0-1). One tile of the distribution corresponds to volume increasing and the other tile to volume reducing.
This result doesn't affect the independent sample T-test because you are analyzing different thinks.
- Badges
- Science topic
- Similar topics
- Biological Science
- Molecular Cell Biology
- Culture Cells
- Cell Line