Conceptual Modeling - Science topic

A bizarre question. Mathematics (in fact, Geophysical models using mathematical approaches) based models are used to predict future climates for different climate regions on Earth. They can be validated based on back casting! Which is seldomly done, by the way! A model altering climates on Earth? That requires more explanation from your side, to actually get a grasp, on what you have in mind? With which boundary conditions? Not clear to me! Is it for you? I don't say it might be impossible, but just imagine what the boundary conditions would be to realize such a venture! We cannot even predict meteorological phenomena a month in advance! Let alone change climates on Earth! Honestly, would you dare to speculate on how this should be brought into the real world?

#NoMercyCV

There are some problems with your concept framework as you hardly able to combine those five factors on the left side of your diagram. The framework also lack of key idea and main contribution. Ask yourself what you are plan to do, and what you wish to contribute? and why it is important. As Phd level thesis, more work has to be done on this framework.

This is only my personal opinion. I hope it can help.

Here are a few popular ones:

1. Lucidchart: Lucidchart is a web-based diagramming tool that offers a free version with limited features. It provides a wide range of shapes, symbols, and templates that can be used to create conceptual frameworks and other types of diagrams. Lucidchart also supports collaboration, allowing multiple users to work on diagrams simultaneously.

2. draw.io: draw.io is an open-source diagramming software that can be used to create conceptual frameworks. It offers a user-friendly interface with various shapes and connectors to design diagrams. draw.io can be used as a web application or downloaded as a desktop application for offline use.

3. Coggle: Coggle is an online mind-mapping and brainstorming tool that can be used to create conceptual frameworks. It allows you to create interconnected nodes and organize your ideas visually. Coggle offers a free version with basic features, and it also supports real-time collaboration.

4. Canva: Canva is a versatile graphic design tool that can be used to create conceptual frameworks and other visual materials. It provides a wide range of templates, shapes, icons, and fonts to design professional-looking diagrams. Canva offers a free plan with many features, and it also has a collaborative feature for team projects.

5. FreeMind: FreeMind is a free and open-source mind-mapping software that can be used to create conceptual frameworks. It offers a simple and intuitive interface for organizing thoughts and ideas visually. FreeMind supports various features like node customization, hyperlinks, and export options.

6. yEd Graph Editor: yEd Graph Editor is a powerful desktop application for creating diagrams, including conceptual frameworks. It provides a wide range of layout algorithms and customization options to create visually appealing diagrams. yEd Graph Editor is available for free and supports various operating systems.

Hope it helps

I like Miro - I think there are lots of other options but in terms of getting started quickly and easily, it's hard to beat.

It's designed as an online whiteboard for collaboration - that you can share with others by sending them a link, (great for online teaching actually) but it's great just as a thinking tool to sort out your own ideas and then organise them visually.

I often screenshot what I've built and use that image in documents etc.

You can create a certain number of whiteboards free, but then need a subscription if you want more. They have a free educational license if you email them and ask - with proof of your educational-user status.

Canva and Figma are alternatives but Miro seems more intuitive to me.

Find it and have a play :-)

Drawing a conceptual model figure can help to represent the relationships between variables in your study visually. Here is an example of how you could draw a conceptual model figure for the study you described:

Here is an example of what the conceptual model figure might look like:

luaCopy code

____

| |

| Y |

|____|

/ \

/ \

/ \

____ ____

| | | |

| X1 |----| X2 |

|____| |____|

| |

| |

| ____/ \____

| | |

|---| M |

|__________|

|

|

|

____

| |

| W |

|____|

Of course, your conceptual model figure's specific layout and design may vary depending on your preferences and the software you use to create it.

Don't forget questions about the theoretical basis of the model and how this model helps to fill the identified gap. It's very important.

Hi! Sir

To validate the model you can not rely only on one data set. Try to apply the model on different data set and find the error.If the error comes lesser in comparison to other model then it will be useful.

Thanks

Thank you, Michael Felix Noel ! All of the PROCESS templates assume the ModV being an IV in it's own right. I wonder whether they apply to my model in which the ModV is a consequence of other antecedents (i.e. the IV, MedV1 and MedV2)?

Thank you, dear Jonathan Samuel

thank you very much

Sorry- hard for me to help without understanding what you are researching.

Dear Bas Bar, I hope that I can shed some light on your negative loadings. It is well-known that the factors in the common factor model cannot be solved for uniquely even if the factor model fits perfectly and the factor loadings are known (see, e.g., Nicewander, 2020; Steiger, 1979, 1996). The unrotated model can be rotated with an infinite number of rotations without altering the model fit. This provides an indeterminacy in the model because no information in the data is available that can determine the best possible rotation for the factors. Therefore, positive or negative factor loadings may be arbitrary.

I once performed an EFA with the result that all items of one of the factors loaded negatively on this factor. When I used Promax rotation instead, all factor loadings were positive.

Nicewander, W. A. (2020). A perspective on the mathematical and psychometric aspects of factor indeterminacy. Multivariate Behavioral Research, 55(6), 825–838. doi: 10.1080/00273171.2019.1684872

Steiger, J.H. (1979). Factor indeterminacy in the 1930's and the 1970's some interesting parallels. Psychometrika 44, 157–167. https://doi.org/10.1007/BF02293967

Steiger, J. H. (1996). Dispelling some myths about factor indeterminacy. Multivariate Behavioral Research, 31(4), 539–550. https://doi.org/10.1207/s15327906mbr3104_7

HTH, Karin

Conceptually a qualitative study cannot have a conceptual framework, though people do it now a days.

Conceptual framework is primarily a part of a quantitative Study.

You can get them in the link mentioned below

48+ SAMPLE Employee Questionnaires in PDF | MS Word | Excel

Koen Van de Moortel

How can we help students develop their competence in constructing and applying, even modifying, internal and external representations? can we advocate integrating formative assessment techniques into modelling practice in the classroom to help teachers and students visualize the change in their thinking through the process of modelling practice? how?

Hello Kirti,

Of course, the model(s) of interest should be driven by your specific research question(s).

It sounds to me as if you could:

1. Evaluate two different models: (a) one with FinLit as a mediator, with some other variable as the target DV (and at least one IV beyond these vars); (b) one with FinLit as a DV;

The only problem with this approach is that you don't have a simultaneous test of whatever the mediation RQ is, coupled with whatever the FinLit = DV RQ might be.

2. Combine the models, so that, FinLit is in fact a target DV for one or more other variables, and it behaves as a mediator in that relationship.

The only problem with the second approach is you'd have to respecify it in at least two other ways to evaluate the mediation hypothesis vs. whatever the hypothesis of interest is when FinLit is the DV.

You'll have to decide which approach makes the most sense given your research aims.

Good luck with your work.

Nishadi Gampathirana Yes AMOS allows running multigroup analysis (MGA) with three or more categories. You can even use Smart-PLS for MGA. A sample paper checking moderation with three categories using Smart-PLS is attached herewith.

Hello Helena Rezende

This topic is very important to me...! (Issues of patient safety/identification, improved care quality and so on...)

I know there's a journal called "Journal of Patient Safety" and one called "Patient Safety".

I know that there is a journal called "BMJ Quality and Safety".. . I am also aware of journals called "Journal of Health Care Risk Management" and "International Journal for Quality in Health Care". I have also heard of one called "Implementation Science"... Would anything like that be suitable?

for global nitrogen data please Ref:

A.Zhang 2021 .

Hello Boris,

If all three variables are measured or observed (vs. latent), then ordinary path analysis or regression could be used to evaluate each hypothesis.

H1: Testing path EI -> EB (for difference from zero, and positive sign)

H2: Testing path SC -> EI (for difference from zero, and positive sign)

H3: Testing path of SC*EI -> EB (note: SC*EI is a product term, which you may wish to form by mean centering each variable before multiplying). For H3, you'd want to include direct link of EI -> EB and possibly that of SC -> EB so as to get the relative contribution of the interaction term compared to the direct influence of the constituent elements on outcome, EB).

This could all be specified in a single path model.

Good luck with your work.

Martijn Aa This design could be conveniently analyzed by using path analysis, which is available in software programs for structural equation modeling (SEM) such as lavaan (in R; free software), Mplus, AMOS, LISREL, EQS, OpenMx, etc.

Path analysis allows you to consider multiple IVs, DVs, and mediators in a single statistical model. It also allows you to conveniently test the indirect (mediated) variable effects for statistical significance. Many SEM programs have specific features for automatically printing and testing indirect effects (e.g., by using bootstrap confidence intervals as provided by, e.g., Mplus).

Dear David, I surely read your answer, but maybe I was not clear. I was supporting your statement that count variables are typically not suitable for OLS regression and hence not for PROCESS. And my second part referrered to that even if the dependent variables were unbounded, metric etc, i.e. ok for OLS regression, PROCESS would not be able to handle the (structural) model that the OP was asking for. PROCESS cannot handle several IVs and several DVs. It is only possible to specify one IV, one DV, two moderators, several mediators and covariables. If you meant this, I think this was not clear and your attachment do not directly refer to this topic.

Best, Rainer

Thank you Pavlo!

Dear Mr. Geiser, dear Mr. Gyamfi I am very greatful for your unexpectedly fast and comprehensive support.

I have done the following two approaches to improve the model fit:

Thus, I now have a sufficiently fit model. According to some influential papers in the research field, CFIs and TLIs > 0.9 are interpretable as good fits.

What would be your advice, what conclusions do you derive from that result, how would you proceed?

The sample has 333 participants, was conducted in multiple countries, the participants have different backgrounds.

Best Regards,

Sven Kovacic

Dear Arivazagan Jay ,

your reasoning is definitely correct in the sense that your methods are indeed able to help you answer your research question. I would however say that if a method is 'good' or 'bad' depends on much more, for instance of your field of research and the level of experience of your intended audience (scientific, practicioners, background in NLU or not...). Since you mention a supervisor but also that you are new to research, I assume you are currently writing some form of undergraduate thesis. If that is correct and you have a page limit to stick to (say 60 pages or less), I would recommand to just select one method. It's always better to perform one analysis thoroghly than to lose your audience half way through your paper.

Good luck with you project!

Prof. Suresh Kumar Govindarajan

In addition to all interesting answers in this thread, you can check the following approach with microseismic, particularly in the case of unconventional reservoirs:

Best Regards.

Also check https://stat.utexas.edu/images/SSC/Site/AMOS_Tutorial.pdf

Framework indicates the components, dimensions, and directions of the study prior to methodological approach of the researcher, while Model is a theoritical concepts that underscore the dimensions and components of empirical investigations.

This is the direct link to various models https://is.theorizeit.org/wiki/Main_Page

A pictorial representation of the interaction of the research variables: dependent, independent and moderating variables.

What you have described is a Structural Equation Model (SEM), where the factors you predict will be assessed by a Confirmatory Factor Analysis (CFA). And yes, you can use Likert-formated items for this type of modeling.

However, eight factors is a high number for this kind of work, and it will generate the need for at least 3 items per factor. That will also create the need for a relatively large sample in order to generate stable estimates for the components in your model.

There are two things to ask first:

If you have a model already designed and you want to find functions/characteristics of some group/object in the literature to map to the model then your study is descriptive. In other words, if you have a defined question/structure to collect data from the literature and describe according to the structure then it is descriptive. For instance, if you want to describe the traits of specific groups according to a pre-planned structure.

Else if your model is not yet conceptualized rather when you will be exploring the literature and find something to convert to a thought or idea then your study is explorative. It means that you have no pre-planned analysis and design structure or model rather you are starting as unplanned. During this exploration you may reach to a new thought and discovered new ideas. In this way, you are clarifying the research further and making a road to descriptive analysis.

Hi Eline Verhaar

R2 values are calculated for endogenous variables only. Your OPB variable is an exogenous variable. Therefore it cannot be calculated.

Best Regards

If your survey was cross-sectional, you will not be able to look into causality. However, you could have use multiple regressions to check what predicts motivation levels in students.

I have this feeling that "causal inference" and "natural-born storytellers" are so different from each other - already at the level of the underlying logic - that it is impossible to lump them both together under a single category. To use the conventional terminology at the moment: on the one hand there is 'evidence-based research' that presupposes the Western 'dualism' between facts and symbols, objectivity and subjectivity, going back at least as far as Descartes - the other is narrative, which presupposes a tale told by an individual person, of the 'this is what I have seen/experienced' type. This is after all the common heritage of the 'West', and most especially the Protestant North of Europe: this acceptance of a fundamental 'difference' between the objective here, the subjective there. From what you write, one has this feeling that what you call 'linearity bias' presupposes that this distinction between objectivity and subjectivity in the above sense no longer holds. Would you be prepared to elaborate on why you think this is so?

Hi Abdullah Aldaeej , I would recommend more technically demanding vector programs such as Adobe Illustrator or Affinity Designer.

I am not sure what you mean by a conceptual model of these vein deposits. Are you referring to epithermal base metal and precious metal veins? See Sillitoe, 2010, Economic Geology, and references therein, for the relation of these veins to intrusive centers and porphyry Cu deposits, based on observations of actual examples. Also, Sillitoe and Hedenquist, 2003, review these types of epithermal veins, again based on the observation of many examples.

The one diode model mimics the I-V relation of one junction solar cells.

The solar cell is a real diode. A real diode has not only the junction voltage drop but also it has an ohmic drop on its body resistance which is the resistance of semicondcutor layers of the solar cell diodes.

As for the current it is found that there is a shunting current components shunting the junction diode of the solar cells. Physically it is microshorts in the junction or at the edges of the junction.

I would like that you see the book chapter:

This book will give you clear insight on the function of the solar cell equivalent circuit elements.

Best wishes

Yves Costa Netto - I think to answer your question, it is important to differentiate the terms "proposition" and "hypothesis". The main difference between the two is that a hypothesis must be testable and measurable, while a proposition deals with pure concepts for which no laboratory test is currently available. A proposition is a tentative and conjectural relationship between constructs that is stated in a declarative form. It is a statement about the concepts that may be judged as true or false if it refers to observable phenomena. When a proposition is formulated for empirical testing, we call it a hypothesis. Since preposition is a declarative statement it does not have to be true, but must be empirically testable using data, so that we can judge whether it is true or false.

The TAM model is used to explain and predict user acceptance of computer technologies. Basically, if an information system can help people do their jobs better, it is perceived as useful. Accordingly, the basic TAM model defines two variables: perceived usefulness and perceived ease of use, in order to quantify users' attitudes towards information technology. In this context, you can check as a TAM example on this link: https://commons.erau.edu/cgi/viewcontent.cgi?article=1354&context=publication

It is about TAM of the Information Systems Security that can help you to respond to your question.

Also, to be confident about the problem, you can use the independent decision variables but they should have significative ponderations. To determine the ponderation, I suggest using an experimental method like "Delphi" as a methodology for constructing your final model.

CFA will give you the total scenario in a single frame. It's more better for describing your model.

Hi Ms. Watermann (Guten Morgen/Tag)

My name is Dr. Etame, I hold a Doctorate in Organizational Leadership from Pepperdine University in Los Angeles-USA. I conducted a qualitative phenomenological study for my study back in 2016/2017. I am currently conducting a sequential explanatory comparative mixed methods study for my Ph.D. in Global Leadership and Change in the same institution. I submitted my draft to my committee a couple of weeks ago and I just received the feedback to start planning for my preliminary oral defense.

To your question, I believe (at least what worked for me) that you should your theoretical framework first as the blueprint that will inform or rather reinforce your methodology as follows:

1- Post positivism (Constructivism and Critical Theory);

2- Axiological, Ontological, and Epistemological underpinnings of the study;

3- Methodology: for Sequential explanatory Mixed-method, consider doing the following:

a- Phase 1: Collect and Analyze Quantitative Data through surveys (closed-ended questionnaires) and test hypotheses.

b- Integrate data by drawing your qualitative sample from your quantitative sample (about 10%)

c- Phase 2: Collect and Analyze qualitative data through semi-structured open-ended interviews.

Once you have set up your conceptual framework, you will have a clear picture of what your research question (s) and hypotheses will be, based on your underpinning theory (ies).

Once you establish this phase, you can now develop your literature review, based on your theoretical framework by funneling down your theory (ies), building your themes and concepts by aligning your variables: Independent Variable (predictor), mediator or moderator (if any), and finally your Dependent Variable (criterion).

I hope this will help unless you plan to conduct a sequential exploratory mixed method.

Good Luck...

In my more than three decades of involvement in hydrologic and hydraulic modeling, I have never heard the term 'simplified conceptual model'. But I have seen models classified or described as conceptual, physically based, deterministic, stochastic, etc. etc.

A common problem in hydrology is that many terms are loosely used without first defining them, which then gives rise to a lot of confusion. This does not happen in physics or mathematics. Hydrologists should try not to use terms that are not well defined.

A hydrodynamic model uses a numerical approximation (e.g. finite element or finite difference methods) to solve a PDE (e.g. shallow water wave equation, Saint-Venant equations). As such, these are physically-based models (or physics-based models if you prefer). You might use the output of an empirical or conceptual model to give the flows at the upper part of a reach being studied (i.e. set a boundary condition for solving the PDE), but a hydrodynamic model is a physically-based model.

Note that I have used "physically-based model" rather than "physical model" - a "physical model" could refer a model built out of concrete, plaster etc. for running simulations in a laboratory (i.e. not in a computer). For example, a scaled model of a floodplain used to study flood characteristics by actually running water over it.

It could be argued that a hydrodynamic model is a "physics-based" model rather than a "physically-based" model.

Better to heading with GPR analysis in exact place, after that able to process with virtual molding.

David L Morgan Yes, I think I did. The conceptual model for evaluating a website consists of three value-based dimensions under which numerous items are included. By testing the validity and stability of the model, I hypothesized the contribution of each item to the users' increased perception of the specific dimensional value.

Thank you very much for your advice. It really helped me go back to basics, which is always important!

May be content analysis of mass media for a certain time interval may help.

@Rmm Pradeep

I am working as a consultant for WRD, Assam. The department receives the model outcomes (prepared by NESAC) during flood time and takes action based on it. The department was a part of the decision makers during formulation of the model and one of the end users too.

When your conceptual framework / research model developed is uniderpinning on previous theoretical framework means you should modify others' theoretical framework by adopting / adapting / add new constructs / variables to form your own / new conceptual framework / research model. When you do that, ensure you complete the following:

CONTROL VARIABLE functions as a means to see how the model responds when certain variables are included or excluded. If you are going to report the result of the study under certain "conditions" and these "conditions" relates to control variables then your supervisor is correct by showing these control variables on the conceptual framework.

NUMBER OF CONTROL VARIABLES of "around 20 CV" is probably too many and is an indication of poor literature review. It is mostly that you pick up these 20 CV from more than one literature source. If so, you should select only what is relevant to your case. Use information criterion to evaluate if you have too many variables. 20 CV may seem like over-fitting the model or just "having no idea" what is relevant.

Multiple regression analysis !

If I understand you correctly, it looks like you're going to need one set of hypotheses for each value that you're going to test, Value1 through Value10. That's a lot of hypotheses!

It looks like you're hypothesizing that both Org Commitment and Job Satisfaction mediate the relationship between ValueX and OCB. This is only meaningful for the values that are significantly correlated to OCB (zero-order correlations). If such a relationship doesn't exist, mediation isn't meaningful.

However, I don't think I would make hypotheses about mediation. I can't imagine that the relationship between valuing hedonism, for example, and job satisfaction is very high (but I might be wrong). However, I think it might be interesting to examine whether holding certain values moderates a) the relationship between between job satisfaction and OCB, and b) the relationship between org commitment and OCB. This would help us better understand these relationships.

No doubt that it depends on your research question(s). However, if you test each aspect in separation actually you are omitting the explanatory power of other aspects which is statistically wrong.

For more detail see....

Halcoussis, D. (2005)

Walter Antolinez got it. We typically use Adobe Illustrator, Leapfrog, or MVS, which are not freeware, to put together graphics like this. I suggest looking at this researchgate thread that covers this topic:

https://www.researchgate.net/post/What_software_do_you_use_to_draw_conceptual_diagrams

Hydro(geo)logic lierature on recharge/discharge processes is predominantly based on experiences observed in humid regions of the world and appears to be biased toward humid understanding of hydrogeoloical phenomenon.

You may consider reviewing the following:

Happy Reading!

Hi Jorick,

unfortunately, including a mediator variable diminishes the (causal) interpretation of the experiment although--of course--mediation is interesting to analyse as it explains why a treatment works (which is the core of the theoretical hypothesis underlying the experiment). This problem also exists in a randomized trial because the mediator and outcome are measured and, thus, influenced by shared unobserved confounders. My 2c:

I would go to a SEM and test the important restriction of T _|_ Y | M, where T is the treatment, Y is the outcome and M is the mediator. In other words: you test the model with the direct effect of T on Y fixed to zero. If this holds, and T has a strong effect (f value >10) than you could support the assumption that M-Y are unconfounded. In this case, T has the role of an instrumental variable. Perhaps you can identify one or more external predictors of M that have no direct effects on Y.

In this paper (sorry for the self-ad), we did that in a randomized scenario experiment:

In that paper you will find the relevant literature regarding. In addition, see the following papers (in descending centrality).

HTH

Holger

Pirlott, A. G., & MacKinnon, D. P. (2016). Design approaches to experimental mediation. Journal of Experimental Social Psychology, 66, 29-38. doi:10.1016/j.jesp.2015.09.012

Imai, K., Keele, L., & Tingley, D. (2010). A general approach to causal mediation analysis. Psychological Methods, 15(4), 309-334.

Imai, K., Tingley, D., & Yamamoto, T. (2013). Experimental designs for identifying causal mechanisms. Journal of the Royal Statistical Society Series a-Statistics in Society, 176(1), 5-51. doi:10.1111/j.1467-985X.2012.01032.x

Eden, D., Stone-Romero, E. F., & Rothstein, H. R. (2015). Synthesizing results of multiple randomized experiments to establish causality in mediation testing. Human Resource Management Review, 1-10. doi:10.1016/j.hrmr.2015.02.001

Gennetian, L. A., Magnuson, K., & Morris, P. A. (2008). From statistical associations to causation: What developmentalists can learn from instrumental variables techniques coupled with experimental data. Developmental Psychology, 44(2), 381-394. doi:10.1037/0012-1649.44.2.381

For modelling and identifying any process, including business capabilities, it is needed to use a systems analysis methodology such as the one I and attaching, developed by me for the concrete case of engineering systems. Firstly you do an external analysis and, as a result of this analysis, you will be able to formulate a conceptual mathematical model (in the most general case it will be multi level and multi objetive, complex model). If the dimension of this model will be large enought you must to decompose it an identify every one of the relations that are part of your model. Once your model is identified you must to develope the optimization procedures for each subtask and the solutions conciliation between them. I hope these ideas will be useful for you

Hi Sam Allen. I'd be interested in hearing/reading more about this. I am conducting mixed methods research into household energy decisions. This includes using qualitative methods, such as interviews and focus groups, to inform the development of context-specific models of household decision making - the model will differ depending on the specific decision households are facing. I will then test the predictive power of the models using experimental methods.

Thank you my colleague.

You can use http://draw.io . it's very flexible and easy to use.

Thanks.

Please read the article available at:

I do not know what is cloud adoption... but I had some research about sustaiability... I guess what can integrate cloud-sustainability-organizations is SERVICE THINKING...

and service thinking recently have found a strong connection to Design Thinking...

Dear Yosef mekonnen Ali

I think yes you can. there are many studies using this model to assessment of climate change.

I agree with you. Probably we need an official/standardized UML Class Diagram notation for ER modelling, but it will take time.

Thank you. But those articles are different.

Thankyou for adding on Celia, I have gone through FMI, and the researches I have gone through including the critical reviews as well, FMI has been suggested to be more suitable and applicable to expert mindfulness practitioners.

Dear Theresa

you have got me hooked. Let me see if I can wrap this up correctly: What you have done is use non-parametric tests (Median test, Mann-Whitney-U-Test) in order to check whether your three latent constructs are distributed equally across groups – instead of independent t-tests – because the Kolgomorov-Smirnoff test told you that you have no normal distribution of your data.

However, I do not see how you have “tested” anything with that. As far as I know, this is (important!) preliminary work for the actual analysis, SEM or regression in your case.

Incorporating your categorical predictor is fairly easy. You simply regress it on the three constructs with the SEM. For that, there must be a linear connection between “interactions” and the three constructs, which can be checked by correlation matrices. If there is no linearity, you probably have to logarithmise “interactions” by creating a new variable (ln(interactions) and check the correlation matrices again.

Given linearity, you would then proceed with regressing “interactions” on the three constructs. The three regression coefficients you get will tell you if, say, a high number of interactions "predict" higher attitudes and purchase intentions.

Lastly, you may then compare means (or medians) and standard deviations of the group with low interactions to the one with high interactions and the three constructs and calculate cohen’s d (https://www.psychometrica.de/effektstaerke.html).

IMPORTANT DISCLAIMER: I am not a statistician by trade and had to learn everything from scratch myself. Since I am not entirely convinced I have expressed myself in a comprehensive way, I could offer you further clarification through Skype or telephone. If you are interested, we could discuss the details by messaging each other directly.

Until then: All the best!

Marcel

Here is a link that may be beneficial:

I'm assuming by error you mean the mass balance or percent discrepancy error reported by MODFLOW in the output file. Check the output file for warnings and errors. You can reduce your convergence criteria in the solver options to reduce this error. PEST will not help reduce this error. You need to figure out why your model is unstable first. Review your conceptual model to see that your model boundary conditions are representative of your understanding of the groundwater system. Good luck.

These articles may help you.

many thanks for your input. All the information is very valid.

Michal

thank you!

Trying to complement what Ganassali presented and focusing on the variables gender and occupation:

I think you can create a model in moderation, where categorical variables could do this moderation between constructs coming from metric variables.

It can get very interesting and reveal other possibilities.

See for example: HAIR, J. F .; HULT, G. T. M .; RINGLE, C. M .; SARSTEDT, M. A Primer on Partial Least Squares Structural Equation Modeling (PLS-SEM). Thousand Oaks. Sage Publications, 201.