Question
Asked 28th Sep, 2017

How do I compare 2 different groups (control vs. treatment) over time? And how do I see at what moment in time they become sign. different?

Hello,
I've got measurements of water quality parameters over time in fish tanks (3 tanks per treatment) and I want to see where in time they, if at all, become significantly different between groups (control vs. treatment). 
I think my setup is like this: my dependent variable is concentration, my independent variable is time, for which I have 3 or more measured time points (three or more related groups). Note: the fish are NOT subjected to both treatments in time. 
And how do I set up my xcel file used for importing data? Do I use all measurements done per group over the time points, or do I use the means per treatment (3 groups in each treatment) as input? 
The results I find by doing Repeated Measures Analysis don't make much sense so I'm not sure about my approach? 
Thank you so much in advance :)

Most recent answer

2nd Nov, 2018
Alanazi Amal
Drexel University
Thank you Amir. I did ANOVA.my question I did the average weight both groups . I got big variance. when I can use average weight instad normal average . what about ifs did not averse weight.

Popular Answers (1)

2nd Oct, 2017
Matheus T. Baumgartner
Universidade Federal de Goiás
Thijs,
The t-test and ANOVA require independence among observations. Since your design includes time, it creates temporal correlations. So, these two options are too much simple. The Repeated Measures ANOVA has an assumption called "Sphericity", which is rarely met. I suggest you an alternative approach. Use Nested ANOVA, with factors nested in this way: Treatment < Tank < Time.
Your excel table would be organized with the columns (from left to right):
Treatment - Tank - Time - Concentration,
Using this design, you will be able to identify if the Treatment differs (which I think is your main objective) and if there was a time at which the treatments become different.
This analysis can be done in R using function aov in the following form:
aov(Concentration ~ Treatment + Treatment:Tank + Tank:Time, data = ___)
6 Recommendations

All Answers (10)

30th Sep, 2017
Jos Feys
KU Leuven
This seems to be a 2 x 3, between x within (repeated measures design); correct me if I'm wrong. If it indeed is a between x within design, just run a two-way ANOVA: group x time. In Excel you would have 6 rows corresponding to 2 groups (control vs. treatment) x 3 tanks (observational units or 'subjects'); there would be 3 columns of concentration data for each of the 3 time points (more columns if more time points).
2nd Oct, 2017
Moshe Gophen
Migal-Scientific Research Institute
Student T0test or ANOVA
1 Recommendation
2nd Oct, 2017
Matheus T. Baumgartner
Universidade Federal de Goiás
Thijs,
The t-test and ANOVA require independence among observations. Since your design includes time, it creates temporal correlations. So, these two options are too much simple. The Repeated Measures ANOVA has an assumption called "Sphericity", which is rarely met. I suggest you an alternative approach. Use Nested ANOVA, with factors nested in this way: Treatment < Tank < Time.
Your excel table would be organized with the columns (from left to right):
Treatment - Tank - Time - Concentration,
Using this design, you will be able to identify if the Treatment differs (which I think is your main objective) and if there was a time at which the treatments become different.
This analysis can be done in R using function aov in the following form:
aov(Concentration ~ Treatment + Treatment:Tank + Tank:Time, data = ___)
6 Recommendations
3rd Oct, 2017
Jos Feys
KU Leuven
As I understand, the analysis suggested by Matheus is a repeated measures ANOVA on the data in 'long' format. The analysis I proposed is the same, but on the data in 'wide' format. With only 3 repeated measures, the 'sphericity' issue is not really a huge problem. With most common statistical software packages (SAS, SPSS, R, STRATA) one can model the covariance structure.
1 Recommendation
4th Oct, 2017
Thijs Gravemaker
Wageningen University & Research
Thank you all for your answers. I think I was able to perform adequate analysis using your tips.
@Jos: it's indeed a 2 x 3, between x withing (repeated measures design). 
I did my statistical analyses like this: 
1.Mixed model ANOVA used to assess whether there were significant differences between and within treatments over time.
2. Separate one-way randomized ANOVA (as follow-up tests) for each time point to assess at what time point these mean values became significantly different between treatments. 
3. Another follow up test, separate one way repeated measures ANOVA's for each treatment to get value for significance within treatments.
5th Oct, 2017
Jos Feys
KU Leuven
This would imply that you predicted and indeed found an interaction group x time. If not, you should consider post hoc tests. In SAS, this can be done on the data in long format using proc mixed, with the statement:
means time / LSD e= time*tank(group);
for the least sign. post hoc test for times.
See the attachment for an example.
6th Oct, 2017
William B. Anderson
University of Waterloo
Instead of a student's T test, try a paired T-test.''
17th Sep, 2018
Alanazi Amal
Drexel University
I have a question about a calculation I am trying to perform on some data I have. I I have two different groups In both groups, each question has different number of students who answered correctly Do I calculate the average weight for the two groups or run a statistic test? I am confused for this point. If I run a statistic test, I will get a small p-value with a small variance. And if I convert every data point to the weight, I get a big p-value because of a big variance
i think ANOVA can help you.

Similar questions and discussions

How do I report the results of a linear mixed models analysis?
Question
45 answers
  • Subina SainiSubina Saini
1) Because I am a novice when it comes to reporting the results of a linear mixed models analysis, how do I report the fixed effect, including including the estimate, confidence interval, and p-value in addition to the size of the random effects. I am not sure how to report these in writing. For example, how do I report the confidence interval in APA format and how do I report the size of the random effects?
2) How do you determine the significance of the size of the random effects (i.e. how do you determine if the size of the random effects is too large and how do you determine the implications of that size)?
3) Our study consisted of 16 participants, 8 of which were assigned a technology with a privacy setting and 8 of which were not assigned a technology with a privacy setting. Survey data was collected weekly. Our fixed effect was whether or not participants were assigned the technology. Our random effects were week (for the 8-week study) and participant. How do I justify using a linear mixed model for this study design? Is it accurate to say that we used a linear mixed model to account for missing data (i.e. non-response; technology issues) and participant-level effects (i.e. how frequently each participant used the technology; differences in technology experience; high variability in each individual participant's responses to survey questions across the 8-week period). Is this a sufficient justification? 
I am very new to mixed models analyses, and I would appreciate some guidance. 

Related Publications

Article
In some applications, measurement errors and multiple repeated measurements often lead to a set-based classification task where objects are represented with a set of samples, and the traditional support vector machines (SVMs) do not work in these settings. To deal with this problem, we construct a new classifier called support function machine (SFM...
Article
The aim of the present study was to investigate the effects of water temperature and pH on total suspended solids (TSS) tolerance of a native fish species, Barbonymus schwanenfeldii (Bleeker, 1854) and an exotic fish species, Oreochromis niloticus (Linnaeus, 1758) in Malaysia. The survival rates of both juvenile fish were tested in different TSS co...
Got a technical question?
Get high-quality answers from experts.