Study Design - Science topic

Yes, a systematic review without a meta-analysis is still valuable. Systematic reviews aim to comprehensively summarize evidence on a specific research question by identifying, evaluating, and synthesizing all relevant studies. They provide a structured approach to reviewing literature, ensuring that the synthesis is based on high-quality, filtered evidence, which helps in reducing bias and increasing reliability

Hello Akuma,

Thank you for posting your question.

Well, the consensus is that cross-sectional studies are neither prospective nor retrospective per se. Rather, they are a snapshot of a population represented by the responses/data collected from the sample.

Some studies may use retrospective or prospective in their manuscripts to better explain the method data was collected.

"Retrospective" cross-sectional studies are based on data that have already been collected previously from databases at a single point in time.

"Prospective" cross-sectional studies, however, are based on data that will be collected in the future through forms/surveys/questionnaires, etc...

In simple words, retrospective cross-sectional studies are based on data that have already been collected (no new data will be collected). Conversely, prospective cross-sectional studies aim to collect new data (i.e, you don't have the data you need when you start the study).

I hope that answers your question!

Namir G Al-Tawil and James Leigh Thank you for your feedback. I agree it is not ideal to use retrospective data as a control group and prospective data to test if an intervention is effective. However, due to low number of participants to recruit from in this case, it almost serves like a quality improvement project but with comparison group being the pre-intervention. A quasi-experimental will make sense using the historical data as the control.

A lot depends on how exploratory your goals are. If there is a well-defined literature in your area, then you should refer to in justifying the new data you are going to collect.

Regardless of that, if the differences between you interviews are small and the reasons for them are obvious, then I don't see any reason to rely on the literature to explain them.

Follow@

I'm a hematologist and hemato-oncologist in training, with skills in statistics!

Hi, I can see a focus/ statistical test being based on two pre- and post-implementaion periods only. However, I would want to base the evaluation on a logic model that links resources, activities and outcomes over time. The transitional period may or may not be necessary for achieving the study's goals, but it might be that things start to improve during the implementation year and tail off during post-implementation. You might want to know if something is immediately successful but cannot be sustained with a different or more focused approach.

Hi Yilak, you can still proceed by making some reasonable estimates. One option in this situation is to rely on data from similar studies conducted in comparable settings or neighboring countries. This helps you approximate the key parameters you need, such as the expected difference between means (effect size) and the standard deviation.

You can then use the "two-sample t-test sample size" formula for comparing two means or an online tool like G*Power to make the calculation.

If there's still uncertainty, consider conducting a small pilot study to get preliminary data on variability. Additionally, it is advisable to adjust the final sample size to account for potential non-response. This can typically be achieved by inflating the sample size by 10-20%, depending on the expected rate of non-participation.

Also, it is important to recognize that differences in sociodemographic and contextual factors can affect the appropriateness of the parameter estimates. Therefore, such estimations should be approached with caution.

Hope this helps!

You may use the APP. "The a priori procedure (APP) is designed as a predata procedure where the goal is to estimate the sample size needed for sample statistics to be good estimates of corresponding population parameters." Please see

Hello, I would agree that predefined inclusion criteria and literature search limits (such as date and English-language) should be defined for a systematic review. Then comes a decision as to whether meta-analysis is possible and if so, which studies should be statistically combined. How similar study populations are clinically and the statistical heterogeniety are drivers to these decisions. One thing that needs more discussion in the community is our current measurement of statistical heterogeneity. According to Michael Borenstein, current measures of heterogeneity, such as I-squared were never actually designed for that purpose. He and others propose the use of the 'prediction interval' for measuring statistical heterogeneity. See Borenstein et al (2021) Introdution to meta-analysis, second edition, Wiley, Chichester.

Aisha Ejura Dahunsi wrote:

1. Repeated measures ANOVA assumes multiple groups or conditions, which is not the case here.

A one-factor repeated measures (RM) ANOVA assumes that there are k = 2 or more measurements of the DV for each independent subject. When k = 2, the F-test from a one-factor RM ANOVA is equivalent to a paired t-test, t² = F. (Try it if you don't believe it!)

The design Ali Itimad described has k = 2 measurements of the DV for each independent subject, but it also has two independent groups of subjects (control & intervention). One option for that design is what I would call a 2x2 mixed design ANOVA*, with Group as a between-Ss factor and Time (pre, post) as a within-Ss (or repeated measures) factor. But another option is ANCOVA, as suggested by Jos Feys (and seconded by me).

* I know that some folks (and maybe some disciplines) refer to any ANOVA model that has at least one RM factor as a repeated measures ANOVA. This is a very imprecise description that can cause a lot of confusion, IMO. I think it is wise to spell out explicitly how many factors there are, including which factors are between-Ss factors and which ones are within-Ss factors. YMMV.

PS- Thank you for acknowledging your oversight in not citing the source of the material you posted earlier. As I have said in other threads, I firmly believe that posting AI-generated content without proper attribution is just another form of academic dishonesty. And I wish people would stop doing it!

This may also help

Thanks, this is a clear and very well laid out set of steps.

I do feel that for non-pharmaceutical studies that removing or minimising the placebo effect may be a disservice to the methodology that is studied.

I believe that the placebo and nocebo effects are integral to understanding treatment outcomes. Some treatments derive their efficacy from enhancing the placebo effect, which is a well-documented and beneficial phenomenon. Conversely, the nocebo effect can exacerbate perceptions of danger and elicit exaggerated responses to perceived threats. The challenge lies in isolating and accurately recording these cognitive influences.

There is a crucial intersection between science and innovation where research should focus on understanding why certain treatments yield positive results, rather than solely aiming to disprove hypotheses through traditional falsifiability methods. Both approaches—proving a treatment's efficacy and attempting to disprove it—can introduce bias.

In my own research, I have found that excluding placebo and nocebo effects might be counterproductive. Instead, we should explore ways to harness and enhance these natural healing phenomena to alleviate chronic pain. Investigating the mechanisms behind treatment efficacy can accelerate the development of effective cures more efficiently than traditional hypothesis testing.

David Sul - Sorry for this late reply, but thank you for the clarification there as well. This makes sense for study design and the anonymous vs confidential is helpful!

You can input one-arm comparisons into RevMan. Simply use the pre-intervention mean and SD for the 'control group' and the post-intervention mean and SD for the 'experimental group'.

If you are including one-arm comparison studies and two-arm comparison studies in the same review, you may wish to do seperate meta-analyses for them, or include them all in one meta-analysis, then afterwards do a sub-group analysis spliting the studies.

Hey! You could use "feminist theory" "political theory" "intersectionality" or "environmental justice" depending on the wording of your research question.

If the exclusion is based on something that happens to participants after randomization, such as not completing the intended course of study, you should retain them in the study for an intention to treat analysis. You can do an RCT without using intention-to-treat analysis but if there are a substantial number of dropouts related to treatment compliance (or similar) your randomly formed groups rapidly become non-random because of something that may be related to your treatment. It introduces bias because the processes leading to dropout will be different in the control group. You must at least acknowledge this limitation and consider the implications.

Using the constant comparative method, as introduced by Strauss and Corbin (1990), outside the context of grounded theory, especially in a multiple case study design, is a creative application that can yield rich insights. While the constant comparative method is often associated with grounded theory, its flexibility allows adaptation to various research designs. Here are some additional thoughts and recommendations:

1. Understanding the Constant Comparative Method:

2. Adaptation to Case Study Design:

3. Literature Review:

4. Methodological Articles:

5. Reflexivity and Transparency:

6. Case Selection:

7. Data Collection Strategies:

8. Coding and Analysis:

9. Triangulation:

10. Flexibility in Analysis:

11. Reporting:

12. Collaborate and Seek Feedback:

Recommended Articles:

These recommendations and considerations should help you navigate the application of the constant comparative method in a multiple case study design. Remember that flexibility and reflexivity are key as you adapt the method to your specific research context.

I think cross sectional study is the best

There are a lot of books about these questions, but here is a page with a very useful resume of it. I hope it will help you.

Davood Omidian Thank you for a detailed answer. I Appreciate it a lot.

I am using the GPower tool, and was wondering which of the MANOVA types are more sufficient?

MANOVA: Repeated measures, between factors

MANOVA: Repeated measures, within factors

MANOVA: Repeated measures, within-between interaction

And for the "Number of Groups" i should pick 6, and the "Number of measurements" i should pick 2 (If i am measuring privacy and view-out)?

The number of measurements does not have anything to do with the 12 different scenarios happening within one group?

Hope I described it clear enough,

Best Regards,

Louis H

Hello Isa,

I suspect you could elicit specific tests (e.g., treat 1 vs. treat 2 for the respective 30-min time point) in whatever software you're using. However, you could always just run (as many as 9) dependent t-tests to do the same thing, with some appropriate adjustment for multiple tests.

The sample size (9 paired score sets) won't offer a lot of statistical power, however, unless the treatment differences are pretty strong, no matter what kind of analysis you elect to use.

Good luck with your work.

This tutorial from Metaboanalyst website describes the methods used for paired analysis: fold analysis, t tests ad volcano plots

Hi Khan,

I'm sharing a paper with you that discusses what you need and points out that there are different variations in the induction of experimental diabetes in animal models between laboratories. My recommendation is that you conduct assays with at least three doses, for example, 50, 75, 100 mg/kg of STZ, and evaluate the glucose levels in your mices. This will allow you to observe the behavior and dynamics, and then determine which model best suits your needs. Also, please consider whether you want to induce type 1 or type 2 diabetes. Just administering STZ will resemble type 1, but if you add diet or nicotinamide, it will resemble type 2 diabetes.

Best regards, Jorge Armando

Please can we have that discussion via email? Kindly furnish me with email address

For such a study, which is an HLA association study you need two cohorts. The first is the patient cohort (Cases) the second is the control cohort which needs to consist of a healthy random group of people from the same ethnic background as the patient group. The sample size depends on the expected effect and the number of comparisons. So it is difficult to say of hand how big the groups need to be.The minimal criteria is 50 patients versus 100 controls (1:2 ratio). If you have less of a chance that you will find any significant result will be very small, this will only be the case if the effect is very large. If you expect only a small difference you need a larger group.

In order to see if your control group is correct and does not contain subpopulations or family members you can check this by using the Hardy-Weinberg(HW) test. Pypop version 7 has such a test.To compare the HLA phenotype frequencies you can use the Woolf-Haldane Odds Ratio with a 95% confidence interval.To test the significance you can use the Fisher’s Exact test(2-tailed).You have to correct for multiple testing and for that, you can use the Sidak’s method or the Bonferroni's method. The correction can be done for the number of associations within a locus.

Hope this may help you.

Thank you.

When someone asks about the clinical significance of a paper before discussing the study design, it may be because they are interested in understanding the potential impact and relevance of the research findings in a clinical context. By evaluating the clinical significance upfront, they can determine whether the study design and methodology are worth exploring further.

Here are a few reasons why someone might inquire about the clinical significance before delving into the study design:

Thank you.

one deals with clearance study of parasite the other deals with impact of this intervention in reduction of case burden

It is entirely dependent on the research questions you are posing. An ethnoarchaeological survey of literature about various groups of people can provide library-based comparative examples from societies with similar social, subsistence, or technological practices, or particular kinds of environments, that are relevant to your research questions. Broad research from the ethnographic literature is always useful for developing a good comparative sense of the problem you are researching. Engaging in a single field-based ethnarchaeological project can be a big undertaking. In general, multiple month longitudinal research is preferable to very short field visits. Almost everything we think about how lifetime hunter-gatherers or small scale agriculturalists do particular things is under-informed , and our ignorance benefits from immersion over time with a particular field site. Setting top field sites is a lot of work, so focusing on a single community develops relationships that are crucial for expanding our learning beyond even the initial research problem. Learning the indigenous language can be critical to more productive investigations, working in translation is not ideal. I have performed ethnoarchaeological research with hunter-gatherers in Venezuela until the country's politics & economy made continued fieldwork impossible (I have 30+ months in the community, learned their language as they were monolingual and spoke no Spanish, and spent over 11 months in permitting in Caracas) and now have over 17 months of ethnoarchaeological experience with traditional Maya farmers in the Yucatan (while many people do speak Spanish, the more May I learn the better my fieldwork). These ethnoarchaeological/ethnographic experiences have been critical for making me a more integrated anthropologist, I now better understand the relationship between archeology, biological anthropology, behavioral ecology, linguistics, kinship studies, etc. in the broader field of anthropology than when I was only an archaeological specialist.

Qualitative researchers are often allergic to the idea of sample size calculation. On the other hand, both from a sampling and from and ethical perspective you want to make sure that the sample doesn't have any important omissions. Whose voices are needed to make sure everyone is heard? So in my opinion it's not the size of the sample so much as the coverage of the study domain.

You can use power analysis to calculate the number of participants needed for a pre-post study design. Here are the general steps:

It is important to note that the sample size calculation is based on assumptions about the effect size, significance level, power, and correlation between the pre-test and post-test scores. These assumptions should be based on previous research or clinical experience to ensure that the study has adequate power to detect meaningful differences.

P: Participant:- Athletes with patellofemoral pain syndrome

I- Intervention:- Not Applicable

C- Comparison:- Healthy match athletes

O- Outcomes:- Lower limb muscle flexibility

Dear Murugesan Arumugam,

Pl refer link (page 22-23) for calculation of loading and maintenance dose based on PK parameters.

Regards

Hitesh Chavda

We can't actually say that there are subtypes of cross-sectional studies but they can be used for descriptive and analytical purposes (even though the analytic purpose is only used to set-up hypothesis that need further investigation)

Hi all,

Adverse events need to be recorded using a safety database (e.g. Argus, safety easy ...) and in a clinical trial setting in the trial database. AEs or SAEs need to be reported to the regulatory authority and to the ethics committee.

The response to the question is different if you are Sponsor, clinical trial centre or CRO.

If you are a member of a sponsor or a CRO, Pharmacovigilance is quite a large domain and I recommend you attend to pharmacovigilance (PV)training.

Here is the link to the PV good practices: https://www.ema.europa.eu/en/human-regulatory/post-authorisation/pharmacovigilance/good-pharmacovigilance-practices

Quite a lot of materials to read and master.

Hello Danielle R. Carnegie The answer is in what you wrote first

"My study design has three factors each with 2 levels for a 2X2X2 within subjects repeated measures anova".

A cohort study is a type of comparative study because it compares two or more groups of individuals based on their exposure to a particular risk factor or intervention. In a cohort study, a group of individuals who are exposed to a particular risk factor or intervention (the "exposed" or "intervention" group) is compared to a group of individuals who are not exposed to that risk factor or intervention (the "unexposed" or "control" group). The goal of a cohort study is to determine if there is an association between the exposure and an outcome of interest.

A non-comparative study, on the other hand, does not involve the comparison of two or more groups of individuals. Instead, it simply describes the characteristics or outcomes of a single group of individuals. One example of a non-comparative study design is a case series, which is a type of study that involves the observation and documentation of a series of cases of a particular health condition or disease. A case series is considered as a non-comparative study since it doesn't have a control group.

A case series is a type of observational study that describes a group of individual cases that have a similar diagnosis or condition. In a case series, the cases are described in terms of their demographic characteristics, symptoms, treatment, and outcomes. The main goal of a case series is to describe the clinical features and the natural course of a disease or condition, and it is a way to identify cases that are rare, unusual or unique to better understand the disease and its impact.

Yes, you can think of a quasi-experimental design as a non-randomized controlled trial. Another name for this kind of design is a non-equivalent control group. There is quite a large literature on quasi-experimental designs.

Bhogaraju Anand Thank you very much. was a great help. :)

Hello Abdi,

The first question is, are you more concerned about: (a) precision of any parameter estimates; or (b) statistical power associated with hypothesis testing resulting from the study?

If (a), then you should consult a text on sampling methods, such as William G. Cochran, Sampling techniques (3rd ed.). Wiley. See this link: https://archive.org/details/Cochran1977SamplingTechniques_201703

If (b), then you'll need to decide: (a) what kind of statistical test makes sense, given the design and nature of score(s) to be collected; (b) how small an effect (which could be a group/treatment difference) you'd like your study to be able to detect, if in fact it exists; and (c) how much risk you are willing to live with of being wrong in your test result, both type I (alpha level) and type II (beta, the arithmetic complement of power). Programs like the freely available G*Power (https://www.psychologie.hhu.de/arbeitsgruppen/allgemeine-psychologie-und-arbeitspsychologie/gpower) can facilitate finding the lower bound sample size which would satisfy your chosen conditions.

Good luck with your work.

The design in any piece of research consists of statements about how the methods (sample, data collection, data analysis, etc.) are linked to the research questions and goals. In contrast, things like grounded theory, phenomenology, etc. are research "approaches" or "traditions." Many qualitative studies do not use one of these approaches, and this particularly likely for studies that rely on semi-structured interviews, followed by thematic analysis.

As for saturation, it is own separate concept related to when further data collection becomes redundant, and thus is seldom related to triangulation.

Also, this is a small "industry" involved in producing lists of criteria for evaluating qualitative research. such as the COREQ and SRQR checklists.

thanks Deborah J Hilton

As my good friend David Morse says life is multivariate but that doesn't necessarily mean that classical multivariate statistical methods are what you need to use. They tend over rely on the multivariate normal distribution. As a example consider 2 group discriminant analysis. Clearly representative of many classic multivariate methods it depends on many classical assumptions while binary logistic regression requirements are less and we know that logistic regression is the optimal method for the classification problem. Multivariate is fine but use the right method which is often not the classical method Best wishes David Booth

There is an important difference between have a single ordinal-scored variable versus a set of ordinal-scored variables that compose a scale. The former requires a non-parametric test, such as Kruskal-Wallis. But when you can combine multiple items into a scale, then you may be able to treat that variable as interval.

Hi,

Your question's last statement itself is the answer. You can refer to STROBE guidelines for the components of such studies:

These Assessment Tools maybe useful

1. CHECKLIST FOR ANALYTICAL CROSS SECTIONAL STUDIES: https://jbi.global/sites/default/files/2020-07/Checklist_for_Analytical_Cross_Sectional_Studies.pdf

2. CHECKLIST FOR CASE CONTROL STUDIES: https://jbi.global/sites/default/files/2020-08/Checklist_for_Case_Control_Studies.pdf

3. CHECKLIST FOR QUASI-EXPERIMENTAL STUDIES (NON-RANDOMIZED

EXPERIMENTAL STUDIES): https://jbi.global/sites/default/files/2020-07/Checklist_for_Quasi-Experimental_Appraisal_Tool.pdf

4. Methodological quality (risk of bias) assessment tools for primary and

secondary medical studies: what are they and which is better?

This is what I would call a very risky and stupid design. Get books on research ethics and clinical trials and try hard to understand what they are telling you. David Booth

I think there should be a "seprate study design" other than cross sectional

How are the sample sizes calculated? I assumed the test is about equal proportions and I get quite different numbers.

But independent of that: the effect size (or proportions) needed are estimates. They are never correct, and they can also be set to some (clinically, economically) relevant values that does not need to correspond to some actual (possibly estimated) effect size.

And above all there is chance involved. Having 0.8 power means that - even if the guessed effect sizes are absolutely correct! - you still have a 20% chance to end up with p>alpha, including all values up to 1. And the 80% chance of having p<alpha includes all values down to 0.

En Español: Lecturología

Study materials are available on it. You may wish to consult them for your understanding

Rigour - an attempt to a research method -https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=&cad=rja&uact=8&ved=2ahUKEwjE4rzsmuH1AhUrzDgGHZsKC-wQFnoECCAQAQ&url=https%3A%2F%2Fbmcmedresmethodol.biomedcentral.com%2Farticles%2F10.1186%2Fs12874-019-0707-y&usg=AOvVaw1c0M3Co8rozyJ7ASOtXSS6

This paper maybe useful

This paper suggests that Inclusive Approach - all study designs including randomized control trials (RCTs), quasi-experimental and observational studies) - is preferable.

Tufanaru, Catalin MD, MPH, MClinSci (EBHC); Munn, Zachary PhD; Stephenson, Matthew PhD; Aromataris, Edoardo PhD Fixed or random effects meta-analysis? Common methodological issues in systematic reviews of effectiveness, International Journal of Evidence-Based Healthcare: September 2015 - Volume 13 - Issue 3 - p 196-207

doi: 10.1097/XEB.0000000000000065

It is cross over. I think better you take two separate groups of students having same IQ level and then teach the same contents to both groups using method A for one group & B for other.

To estimate a prevalence (P), the sample size is: n = z²PQ/u². With Q= 1-P, u = required precision, z = 1.96 for α = 5% . If several estimates of P are advanced, the one that generates the largest sample size is selected. At the same required precision and α error , the estimated sample size is the greater the closer P is to 50%. If we have no idea of P, we can adopt the most unfavourable situation where P = Q = 0.50. Hence: n = z²/4u². For example, for an required precision of 0.020 (2.0%), at risk 5%, the sample size to be taken is: n = 1.96² / (4 * 0.02²) = 2401.

Thank you @Ronán Michael Conroy for correcting me. Appreciated

Dear Matthias Luethi

Generally, Managing machine learning experiments, trials, jobs and metadata using Amazon SageMaker

Kind Regards

It is difficult to answer this question without more information. In particular are the patients being randomly assigned to either a novel treatment or to standard care or placebo? If the trial is placebo controlled is it blinded?— in other words are the patients and the assessor unaware of which group they have been assigned to. I’m sure people would help you define your study type if you could provide more information

Thanks all of you for your kind answers. I got it because of your support. Thank you guys ,have a blessed life

Hello! you might want to check this free online block randomization with random block sizes app. With this randomization method you can ensure an even shuffle between visual cues. https://sigdaan.com/randomization/app/randomization-app#

A cross-sectional study is the study design of choice to determine the prevalence since it is not costly to perform and does not require a lot of time and captures a specific point in time

If you are doing a study to see if pes planus is a risk factor for knee OA, then you are in the territory of a case-control study. What you have to show is that pes planus is more common in patients with knee OA than it is in people (probably of a similar age) from the same population who do not have knee OA.

Just studying the patient population will give you the prevalence of pes planus but it won't tell you if this prevalence is unexpectedly high or low. To know that, you need to establish the prevalence of pes planus in a group from the same population who do not have knee OA.

Does that make sense?

Hello Ina,

As an observation, 23 cases, split into two batches, isn't a lot for yielding stable estimates of regression lines in the first place.

You can execute the steps for a Johnson-Neyman analysis in spss (or pretty much any statistical software package). It's not, however, a direct menu option...you'll have to use syntax.

Here's a link that outlines the process: https://www.glmj.org/archives/articles/Ji_v42n1.pdf

If you'd rather try your hand using the freely available R system, here's a link that walks you through the steps: https://kenstoyama.wordpress.com/2018/01/21/the-johnson-neyman-tecnique-and-an-r-script-to-apply-it/

Good luck with your work.

Hello Emilia,

With 10 individual DVs representing different domains, it doesn't seem wise to incorporate all of them into a single analysis (maybe three: physical measures, emotional measures, motivation). As well, what manova/mancova does is generate linear combinations of the DVs that maximally differentiate group means (centroids in space). The problem is, figuring out what those linear combinations (e.g., how each DV was weighted in the combo) mean with respect to group differences. It can be done, of course, and using (descriptive) discriminant analysis is one way to focus on the (multivariate) nature of the differences in groups on the set of measures.

The related issue is that manova/mancova is less powerful than anova/ancova (paradoxically, it can be more sensitive in that it takes into account the relationships among the variables). So, to achieve a given power takes a larger sample size for the multivariate test than the univariate test, all other things equal.

That said, it would probably be easier to just run a set of ancovas (or the equivalent regression models). You'd likely want to adjust for experiment-wise alpha level (with 10 tests in all). The only drawback to the univariate perspective is that you won't see profile differences, which may be important in addressing your specific research questions.

If assumptions are a concern, many software packages offer either: (a) bootstrap or resampling estimates; (b) "robust" tests' and/or (c) exact tests. Both (a) and (c) allow you the luxury of not having to worry about distributional shape or variance equality.

Good luck with your work.

Hi Erin,

Yes, you have enough participants to complete the study as it stands. Just ensure when reporting the results you don't overgeneralize and you'll be fine.

Best of luck!

Thematic Analysis or Grounded Theory Analysis principles may be incoporated

Ivan Karuc (MINORS) tool or EPOC RoB tool

Good luck

retrospective cohort study,as you went back to study relationship between exposure and outcome

I agree with what Jesper said, It can be a prospective cohort study, from which the association between dependent (outcome) variables with other independent variables (exposures) can be calculated as in any prospective study design. However, the design of the study will depend much on your research question.

Hello Sili,

Wow; it seems as if monitoring the lifespan of each individual animal in the study would have been a basic expectation for making longevity comparisons. As well, do I understand correctly that you wish to evaluate the impact of (or on) 200 variables as they might or might not relate to average lifespan for a group? If so, then the study seems woefully undersized in order to be able to do that with any precision. Just the thought of running 200 tests means that the expected number of Type I errors would likely be high enough to worry about, unless you set the per-comparison significance level very low.

What this leaves you with as options is not really an attractive pool.

1. You could run correlations of mean levels of a variable with the mean lifespan for a group. Unfortunately, that's 5 pairs of data points, which is not going to lead you to a very precise estimate.

2. If you have measurements for each animal on a given variable, you could run a one-way anova (using the 5 groups as the IV). Again, these won't be very powerful tests unless the effects are pretty large. The sheer number of such tests also needs to be considered, due to the aggregate Type I error risk. I would not suggest manova, as the number of cases is simply too low to afford statistical power.

3. I would not recommend the OLS regression method you outlined, as you would be artificially masking variation in lifespan.

4. Your ability to compute and report ORs (odds ratios) for survival is severely hampered by not knowing whether a given group 3 animal did or did not outlive a given control group animal. Presuming they all did may well distort the reality of the outcomes, given the description of what you know. If you have, by hour / day /week / month the number of surviving cases in each group, you could try a proportional hazards analysis, but the sample sizes are likely far too low to offer much precision here (also, it sounded like you didn't really have this type of information available).

5. Just report the average lifespan of each group, the mean of each group on each variable (if individual animal results are available, then you should include CIs as well), call the study and presentation "exploratory," and leave it at that (with the obligatory recommendation that, "further research is needed").

Good luck with your work.

https://doi.org/10.1016/j.arcmed.2021.05.002

Thank you😊😊😊

Multivariate analysis of variance (MANOVA) is used to assess multiple dependent variables (DVs) concurrently. MANOVA is an extension of the analysis of variance (ANOVA), which is used for only one DV. The following could be a good read.

Frost, J. (2018, November 13). Multivariate ANOVA (MANOVA) benefits and when to use it. Statistics By Jim. https://statisticsbyjim.com/anova/multivariate-anova-manova-benefits-use/

Good luck,

Robert Boer, Mohanad Kamaleldin Mahmoud Ibrahim Babak Jamshidi Thank you everyone for answering. Robert Boer I'm currently conducting a systematic review on the availability of TB services pre- and post- pandemic. I've included that report in my review, and I need to know the study design for reporting on study characteristics and study quality assessment. May I ask how would I assess the study quality for a surveillance study? This is my first time conducting a systematic review and there are a few things I'm still unsure of, any advice would be highly appreciated.