POPOOLA, S. Modifikace regresní funkce [online]. Brno: Vysoké učení technické v Brně. Fakulta strojního inženýrství. 2022.

Posudek vedoucího

Žák, Libor

The diploma thesis Modification of Regression Function deals with an angled regression function. The assignment of the diploma thesis is a reaction to COVID19, where the increase and decrease of positive cases has an exponential shape. The aim of the diploma thesis was to convert the data into a linear regression model by using transformation. Find the appropriate description with the help of linear regression and convert the results back. The diploma thesis consists of a theoretical part, a description of the problem, finding a solution for discontinuous and continuous regression functions. Then apply these solutions to the entered data (COVID19). Working with a student was very difficult. The main reason is his ignorance of basic knowledge corresponding to the undergraduate degree. When solving a problem, he guesses the result rather than deriving it. If I recommended him to correct an expression at work, he removed it instead of correcting it (eg bad derivation). Therefore, the thesis submitted by him is the fourth version of the diploma thesis, which I check. The diploma thesis is inconsistent (different designations for the same terms), some derivations are missing, others are twice. It is full of mistakes and typos, it replaces linear regression with a regression line, etc. The data is displayed only as a picture of the Excel table. As well as some results. The pictures are not uniform. Listing all the shortcomings would be very comprehensive. It was also clear from our meeting that he did not understand some parts of the text and the derivation of relationships. Working with literature is also bad, where, for example, on page 33 there is a text that I sent him (as a guide to the task, what to do) and which is assigned under literature [40]. I do not recommend diploma thesis to defend and I evaluate failed.

Posudek oponenta

Hübnerová, Zuzana

The thesis aims at the simplest problem of piecewise linear regression model. While reading the thesis I learned a work „haphazardly“ and I can use it directly in my review. Mr. Popoola copies different texts from various websites (with referring to them) haphazardly thus the chapter 1 is an inconsistent mixture of statements and definitions connected with linear regression. Moreover, many topics included in the introduction chapter on regression are not within the scope of the thesis. It is not clear why chapters 1.6 and 1.7 are not joint as they cover the same topic. There are many mistakes and incorrect terms, which I noted, some of them are listed after my conclusion. However, the text is flooded with them. Moreover, a change-point test is promised in the abstract, but I did not find any. I STRONGLY suggest the rejection of the thesis! - Typos in keywords - Linear regression does not necessarily mean that the dependence of the explanatory and explained variable needs to be linear. P5 - p. 7 1.2.2 the error term is missing - p. 9 paragraph on Heteroscedasticity repeats. - p.10 parameter beta is called obscure boundaries. - p. 11 it is not clear what H stands for - p. 12 it is not clear how section 1.5 implies section 1.6, the formula 1.6.1 especially - p. 14 comments on inversion of X’X repeat. - p. 20 It is not explained what S_A stands for. - The equation after (1.6.17) is not derived, nor explained. - p. 20 The terms “a populace with obscure mean and fluctuation “ are not correct - The statement below (1.8.2) is incorrect. - Model, where „Variance of X can sometimes be a random variable.“ Is commented on but not described appropriately. - Unlike a standard notation used in the literature, hat beta is used for a standardized regression parameter. Later on hat stands for estimation - 1.12.2 The confidence intervals are given without the assumption on normality of the residuals - Presenting data as in Figure 2.1 is ridiculous - (2.4.1) the error term is missing - Step from (2.4.4) to (2.4.5) not correct - 2.4.5 Comments on conditions of the correlation coefficient are not mentioned. Nor is the standard approach to calculating r given. - 2.4.6 T-test is not given in the introductory part on regression nor is it referred to a source. - 2.5. Comments on piecewise linear regression best fits, without stating the piecewise LRM at all - 3.1.2 It is not clear how many observations there are in the first and second part of the sample. - Statement after (3.1.6) is not correct. - The statement that matrix tilde A is positive definite does not mean it is positive. - (3.1.13) it is not clear what phi stands for - (3.1.15) not correct

eVSKP id 137283