BookmarkSubscribeRSS Feed
🔒 This topic is solved and locked. Need further help from the community? Please sign in and ask a new question.
Quartz | Level 8

## Could we consider the cell experiment at multiply time points as a repeated measurement design ?

Hi all !

We conduct a study for exploring the mechanism of osteoporosis. There is an experiment about observing the biological reactions of osteocyte when we treat it as three drugs (A, B, C type).

First, we have isolated the osteocyte from tissue.

Second, we will perform cell culture and divide them into five groups (I, II, III, IV, V).

Third, each cell group (I, II, III, IV, V) will also divide three groups for treating with the three drugs (A, B, C type).

We will treat the osteocyte with the three drugs at five different time points ( I group (A, B, C type) at 12 hours, II group (A, B, C type) at 24 hours, III group (A, B, C type) at 36 hours, IV group (A, B, C type) at 48 hours, V group (A, B, C type) at 72 hours).

(In fact, we divide the cell for 5X3=15 groups.)

In our experiment, the primary outcome measure for the biological reactions was continuous variable. And we detect the primary outcome measure at the five time points, respectively.

Thus, the question is: should we consider it as a repeated measurement design?

Although the five groups cell was originated from the same primary cell for the same mice, we think the cell experiment is different from the human or the animal.

The five groups cell at five different time points might satisfy the independence requirement. So, we should analysis the data by the ANOVA at five time points, respectively, rather than Repeated measures ANOVA or Linear mixed model or GEE.

Is right? Thanks a lot!

1 ACCEPTED SOLUTION

Accepted Solutions
Jade | Level 19

## Re: Could we consider the cell experiment at multiply time points as a repeated measurement design ?

I agree that this is not a repeated measures in time design.  The experimental unit is the aliquot (out of five) that is treated with a single drug at a given time.  What you have is a randomized block design, with each time by source measure as the exptl unit, and time and drug (and time by drug interaction) as fixed effects.  The question then becomes how to handle cell groups I to V, from each of which 15 samples were taken.  I would probably consider cell group to be a random effect.  A mixed model approach would work well here, with the fixed effects previously listed and cell_group as a random effect.  The means obtained are then conditional on the cell_groups.  A GEE approach could also be done instead, to get marginal estimates.

SteveDenham

3 REPLIES 3
Jade | Level 19

## Re: Could we consider the cell experiment at multiply time points as a repeated measurement design ?

I would consider this design as a split-plot in time, best analyzed as a repeated measures analysis.  The whole plot has five replications from each osteocyte parent sample.  As time goes by, you split the plot into 3 subplots (drug A, drug B, drug C), giving 3 treatments at each time point.  GEE or linear mixed model would be my personal choices, depending on your need for marginal vs. conditional estimates, and the inference space you need to answer your question.

SteveDenham

Quartz | Level 8

## Re: Could we consider the cell experiment at multiply time points as a repeated measurement design ?

Prof. SteveDenham, Thanks a lot ! But, if we expect those five groups just to be technical replicates, separate but otherwise equivalent samples from the originating cell culture, then there's no need to treat this as a repeated-measure design. Is right?
Jade | Level 19

## Re: Could we consider the cell experiment at multiply time points as a repeated measurement design ?

I agree that this is not a repeated measures in time design.  The experimental unit is the aliquot (out of five) that is treated with a single drug at a given time.  What you have is a randomized block design, with each time by source measure as the exptl unit, and time and drug (and time by drug interaction) as fixed effects.  The question then becomes how to handle cell groups I to V, from each of which 15 samples were taken.  I would probably consider cell group to be a random effect.  A mixed model approach would work well here, with the fixed effects previously listed and cell_group as a random effect.  The means obtained are then conditional on the cell_groups.  A GEE approach could also be done instead, to get marginal estimates.

SteveDenham

Discussion stats
• 3 replies
• 392 views
• 1 like
• 2 in conversation