I am comparing the number flowering in normal plants and 3 mutants. I have 60 normal plants and 60 of each mutant plants. The flowering were measured at 6 consecutive weeks and the data is like the following:

normal <- c(0, 5, 30, 55, 58, 60)
mutant_A <- c(0, 0, 10, 25, 45, 50)
mutant_B <- c(2, 10, 50, 59, 60, 60)
mutant_C <- c(0, 4, 28, 56, 59, 59)

I was wondering what the best way to analyze the data is. If I do a repeated measure ANOVA, it seems not considering the accumulative nature of the data (i.e., once a plant is flowering, it will never be counted as non-flowering. Hence the data for each type of plants only go up or keep the same.).

One possibility I am thinking is use the number of "not flowering" and do a survival analysis? Is that a reasonable idea?



Survival analysis is a good idea for your case. Six time points might be too short, though...

For a proper analysis, your data needs to be in a "longer" format, where you have 4*60 rows (one for each plant), and four columns: the first is the factor (WT, A, B, C), the second the "left" time (time of inclusion in the study, 0 for all your plants), the third the "right" time (the week in which you observed the plant having flowered or if it didn't flower, the end time point of the study: 6), and the fourth finally the censoring status (0-1, has the plant flowered by the last time point).

Also you need to take care, the data is interval-censored: you only observe at a certain time point if this plant flowered since the last time point.

  • 1
    $\begingroup$ Good catch with the interval censoring. I think the longer format is unnecessary: one could use weights instead. The final results should be the same in either case. $\endgroup$ – Cam.Davidson.Pilon Jun 20 at 23:20
  • $\begingroup$ Hi Edgar and Cam.Davidson.Pilon, is there any other way to analyze this type of data besides using survival analysis methods? I agree that the data points are not enough and interval is not continuous. Thanks! $\endgroup$ – l0110 Jun 21 at 12:11
  • 1
    $\begingroup$ Survival analysis based on Kaplan Meier curves / proportional hazards model should be fine with the number of data points and is especially suited for discreet time points data. $\endgroup$ – Edgar Jun 21 at 13:09

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.