Hello, I conducted a pilot study with limited number of participants n=6. originally planned 20, but due to covid got less participants. I was checking A1C before and after telephone-based intervention (two tailed t-test).
For this small sample should I even check the normality?
Also I did not see any statistical significance, also probably to the small sample size. Should I use descriptive statistics only?
As you have small sample, neither the shape of the error distribution nor the presence of outliers can be judged with reasonable confidence.
Therefore, you can check two things; one, what literature contains about the normality of the variable under consideration, and two, look at the descriptive statistics, namely, mean, median, mode, range, and quartile deviation.
Hello.
A few comments:
1. You can focus on effect sizes rather than statistical significance.
2. There are adjustments for small sample bias; for example, for the SD. So, maybe there is some adjustment for the type of analysis you want to do.
None of the above options are optimal, but they are reasonable, and possibly unavoidable.
Normality is highly essential when small samples are dealt. However, non parametric tests can be applied if data normality assumption failed.
Many times when we are carrying out a study in relatively new field or in a new area, we often go first for a pilot survey. The idea of a pilot survey is to form some idea which otherwise is not known to the investigator. Let us say that one researcher is interested in studying the opium consumption among rural areas of Jaisalmer district of India. To decide about the possible sample size, you need some information about the prevalence of opium use among the population so that a optimum sample size could be decided. Next, we do not know the questionnaire which we want to use for the survey is feasible or not? A pilot survey carried out with possible questionnaire would throw some light on the type of answers to be received and the likely problems faced in the administration of the framed questionnaire. So, in short, a pilot survey can be conducted with different purpose in mind. According to me, I suggest a pilot survey for the following purposes:
1) Determination of likely prevalence of opium among the study population so that an optimum sample size could be decided?
2) To test the validity of the questionnaire framed and likely problems faced in the administration of the questionnaire?
3) Time taken to complete a questionnaire so that a proper time schedule can be formulated for completion of the scheduled survey?
4) Likely non-response rate so that the sample size can be raised accordingly. For example, in general health survey, 25% non-response rate is common.
5) Pilot survey improves the clarity in the mind of the investigator about his survey, questionnaire, time schedule etc.,
So, I am not clear why did you do the pilot survey? Where is the question of compromise in the pilot survey results itself? I will not agree with small sample in the case of a pilot survey.
Please try to complete the appropriate sample size. Analyze it. The results obtained from appropriate sample size will give you valid results otherwise you may face some problems with final analysis of the data.
Ina Vate , with such a small sample size, it is not possible to use traditional (frequentist) statistical methods, as these methods are asymptotic (i.e. valid for large samples). I would suggest to use Bayesian methods, as these methods are valid even with very low small samples.
Here you can use a dependent t-test for small sample size data. The t-test can be applied for the extremely small sample size (n = 2 to 5) provided the effect size is large and data follows the t-test assumptions. For dependent t-test, it is advisable to have a high within-pair correlation (r > 0.8) to get a high statistical power (>80%) for small sample size data.
Read more here
https://www.reneshbedre.com/blog/ttest.html
- Badges
- Science topic
- Similar topics
- Phytochemistry
- Extracts