ISI Digital Commons (Indian Statistical Institute )
Not a member yet
7571 research outputs found
Sort by
Essays on Corruption: The Role of Information, Beliefs and Incentives
No full textIt is well acknowledged that corruption is rampant in low-income countries. However, there is a less than ideal propensity to take action against it. Lack of information is one important factor that might explain why citizens don’t take an initiative to fight against it. Another important issue is citizens’ ability to co-ordinate with each other to reduce corruption. This implies that when individuals decide to take an action, they need to have some knowledge about whether and how others are going to act, since the success or failure of many anti-corruption efforts depend on individual actors being able to co-ordinate. The first essay addresses the importance of these two channels in context of anti-corruption actions. Building on the same context, the second essay posits that the drive to take actions against corruption might be strong when major crisis or disaster is fresh in the memory, thereby making it more personal. A period of crisis heightens public attention- a fact that is not lost on politicians / public officials. The third essay explores the delivery of a public good in the context of a period when there is heightened public attention during the electoral term of an incumbent politician. Anticipating such behavior from the public, politicians might time their actions in a way that would be more rewarding or to their advantage. Through the third essay, we empirically test if such a manipulation can be detected in the provision of an important public good, both in terms of its quantity and quality. In the first essay, we conduct an online experiment to test whether increasing awareness of corrupt practices and/or updating beliefs about others willingness to take action against corruption affects individuals’ own anti-corruption efforts in the health sector during the ongoingCOVID-19 pandemic. Subjects from India are randomized into three treatment groups. In the first treatment, subjects are exposed to increased awareness about corruption. In the second,we correct their misaligned beliefs about others’ willingness to stand up against corruption and in the third treatment, subjects are exposed to both increased awareness and belief correction. Within each treatment group we randomly assign subjects to different anti-corruption actions that vary in their private costs and expected benefits. Our results indicate that our treatments’ impact on subjects’ personal decision to act depends on the relative costs and benefits of anti-corruption actions. In the second essay, we exploit the unexpected occurrence of a health crisis to answer if critical junctures drive citizens’ motivation to fight corruption. We elicit perceptions about corruption in the health sector and the willingness to act against it in an online survey, conducted with nearly 900 men during the height of the second wave of the COVID-19 pandemic in India between March and July 2021. We assess how these measures changed with the severity of the pandemic during this period, using both real-effort and hypothetical measures of citizen activism. We find a significant surge in the proportion of respondents agreeing to participate in protests after the COVID-19 peak, as well as in the willingness to take anti-corruption actions. Furthermore, we observe a substantial increase in subjects’ perception of corruption and their level of information on citizen rights and entitlements during the same period. The evidence,therefore, suggests that the second wave of the pandemic not only acted as a focal point leading to greater willingness to act, but it also increased the probability of citizens taking an anti-corruption action. In the final essay, we analyze the incentive of politicians to engage in corrupt behavior- specifically through their predisposition to adjust policies in systematic ways - around elections in India. We leverage a nationwide road program covering 150,000 roads from 18 large Indian states to demonstrate the presence of election cycle progressively through different stages of program implementation over two decades. Through heterogeneity analysis, we document that politicians are more likely to increase project sanctions in areas with low literacy, but not subsequently improve project award and completion. Additionally, re-election chances are positively correlated with increase in project awards, but tighter electoral competition is unlikely to explain the continuance of electoral cycle in this program; rather, it acts as a force of scrutiny, likely increasing the efficiency of road delivery around elections
Large Sample Inference in Finite Population Problems
No full textIn sample survey, estimation of different finite population parameters like, mean, median, variance, coefficient of variation, correlation and regression coefficients, interquartile range, measure of skewness, etc. was considered extensively in the past. However, comparison of different estimators of the same parameters has been limited. Also, asymptotic theory for several estimators has not been adequately developed. One of the main objectives of this Ph.D. thesis is to compare various well-known estimators of finite population parameters under different sampling designs based on their asymptotic distributions. Another objective of this thesis is to understand the role of auxiliary information in the implementation of different sampling designs and in the construction of different estimators. Four different chapters in this thesis focus on four major topics. In the second chapter, several well-known estimators of the finite population mean and its functions are compared under some commonly used sampling designs. A similar comparison is carried out in the third chapter for the case of mean, when the data are infinite dimensional in nature. In the fourth chapter, the weak convergence of different quantile processes are shown under several sampling designs and superpopulation distributions, and these results are used to study asymptotic properties of estimators of various finite population parameters. Finally, in the fifth chapter, the asymptotic behaviour of the estimators obtained from different regression methods are studied in the context of sample surve
Design and analysis of MDS and Near-MDS Matrices and their application to lightweight cryptography
No full textwe focus on studying MDS and Near-MDS (NMDS) matrices and explore their construction in both recursive and nonrecursive settings. We present several theoretical results and analyze the hardware efficiency of MDS and NMDS matrix constructions. We begin by providing a comprehensive study of MDS matrices over finite fields. This study not only summarizes existing results but also reveals deep and nontrivial connections among various constructions of MDS matrices. Next, we delve into the study of various sparse matrix structures for the construction of both MDS and NMDS matrices in recursive settings. Additionally, we explore various structures for the nonrecursive construction of NMDS matrices, including circulant and left-circulant matrices, as well as their generalizations such as Toeplitz and Hankel matrices. Whenever possible, we also make comparisons between the results of NMDS and MDS matrices. Next, we present various techniques for direct constructions of MDS and NMDS matrices in both recursive and nonrecursive approaches. In the recursive approach, we derive recursive MDS and NMDS matrices from companion matrices, while direct constructions of nonrecursive MDS and NMDS matrices are obtained by using two generalized Vandermonde matrices. Furthermore, we propose a direct method for constructing involutory MDS and NMDS matrices. Finally, we introduce FUTURE, a new SPN-based lightweight block cipher designed with minimal latency and low hardware implementation cost in mind. To achieve the best diffusion in the linear layer, FUTURE incorporates an MDS matrix in its round function. While the use of MDS matrices in lightweight block ciphers is typically avoided due to their high implementation cost. The MDS matrix in FUTURE is composed of four sparse matrices, striking a balance between diffusion property and implementation cost. In addition, FUTURE adopts a lightweight yet cryptographically significant Sbox, which is formed by combining four different Sboxes. By combining these design choices, FUTURE successfully combines lightweight implementation with the desirable properties of MDS matrices, offering an effective solution for designing lightweight block cipher
Quantum symmetry in multigraphs and its applications in physical models
No full textWe introduce various notions of quantum symmetry in a directed or undi- rected multigraph with no isolated vertex and explore relations among them. If the multigraph is simple (with or without loops), all our notions of quantum symmetry reduce to the already existing notions of quantum symmetry provided by Bichon and Banica. Using the machinery developed, we have also provided applications of quantum symmetry in some mathematical and physical models
Privacy Aware Machine Learning
No full textPrivacy preserving computation is of utmost importance in a cloud computing environ- ment where a client often requires to send sensitive data to servers, offering computing services, for computational purposes over untrusted networks. Sharing the raw or an ab- stract representation of a labelled or unlabelled dataset on cloud platforms can potentially expose sensitive information of the data to an adversary, e.g., in the case of an emotion classification task from text, an adversary-agnostic abstract representation of the text data may eventually lead an adversary to identify the demographics of the authors, such as their gender and age, etc. The leakage of sensitive information from the data may take place due to eavesdropping over the network or malware residing at the server. Privacy preserving computation workflows aim to prevent such leakage of sensitive information by introducing a suitable encoding transformation on sample data points. Such an encoding strategy has dual objectives, the first being that it should be difficult to reconstruct the original data in the absence of any knowledge of the encoding strategy and its parameters. Secondly, the computational results obtained using the encoded data should not be substantially different from those obtained using the same data in its original form. Standard encoding mechanisms, such as locality sensitive hashing (LSH), caters to the first objective of privacy preserving computation workflow, the second objective may not always be adequately satisfied. In this thesis, we focus on the second objective and the computational activity that we focus on is a supervised classification task in addition to the K-means clustering, which has been widely used for various data mining jobs. Here, we have addressed the problem of privacy preserving computation on the above two tasks in three different ways, Initially, we have proposed a new variant of the K-means algorithm which is capable of privacy preservation in the sense that it takes binary encoded data as input, and does not require access to the data in its original form at any stage of the computation. The proposed strategy is capable of producing the required number of clusters which are sufficiently close to the respective clusters computed from the original non-encoded data. The results of the proposed strategy on image or text data are either comparable or outperform the standard K-means clustering algorithm. Secondly, we have explored a deep metric learning approach to learn a parameterized encoding transformation with an objective of maximizing the alignment of the clusters obtained in the encoded space with the same obtained from the original data. To this end, we train a weakly supervised deep network using triplets constructed from the output of a clustering algorithm on a subset of the non-encoded data. Our proposed method of weakly- supervised approach yields more effective encoding in comparison to approaches where the encoding process is agnostic of the clustering objective. Finally, we propose a universal defense mechanism against malicious attempts of stealing sensitive information from data shared on cloud platforms. More specifically, our proposed method employs an informative subspace based multi-objective approach to produce a sensitive information aware encoding of the data representation. A number of experiments conducted on both standard text and image datasets demonstrate the ability of our proposed approach to reduce the effectiveness of the adversarial task without remarkably affecting the effectiveness of the primary task itself
Dimensionality Reduction for Data Visualization and Classification
No full textIn this thesis, we identify a few gaps in the existing methods of dimensionality reduction for data visualization and classification and propose some solutions to those as summarized below. Most of the data visualization methods do not learn any explicit function to project high dimensional data to a lower dimension. To overcome the difficulty associated with the absence of an explicit map, in Chapter 2, we propose a framework to estimate explicit maps for data visualization in a supervised setting. The quality of output of any regression-type system depends on the quality of the target data. However, even for simple data, sometimes the target data for visualization may be severely distorted. We present a framework that can significantly correct such distortions in the output for data visualization. For any supervised data visualization method the availability of target data is indispensable, which limits the applicability of such methods. Another problem with most of the methods is that they always produce some output given any input, even when the test input is far from the “sampling window” of the training data. In Chapter 3, using a fuzzy rule-based system (FRBS), we propose an unsupervised approach to learn explicit maps for data visualization that addresses the previously mentioned issues. The proposed method can project out-of-sample instances in a straightforward manner. It can also refuse to project an out-of-sample instance when it is far away from the sampling window of the training data. We have demonstrated the generality of the proposed framework using different objective functions for learning the FRBS. When a data set has significant differences between its class and cluster structure, features selected considering only the discrimination between classes would lead to poor clustering performance. Similarly, features selected considering only the preservation of cluster structures would lead to poor classification performance. To address this issue, in Chapter 4, we propose a neural network-based feature selection method that focuses both on class discrimination and structure preservation. For large datasets, to reduce the computational overhead we propose an effective sample-based method. When a data set has class-specific characteristics, selecting a single feature subset for the entire data set may not characterize the data correctly, although the classifier performance may be satisfactory. To address this, in Chapter 5, we have proposed class-specific feature selection (CSFS) schemes using feature modulators embedded in a fuzzy rule-based classifier. The parameters of the modulators are tuned by minimizing a loss function comprising classification error and a regularizer to make the modulators completely select or reject features in a class-specific manner. Our method is free from the hazards of most of the existing CSFS methods, which suffer due to the use of onevs- all strategy. We have extended the CSFS scheme so that it can monitor class-specific redundancy between selected features. We note here that data from a particular class may have multiple clusters and different clusters may be effectively defined by different subsets of features. To address this, finally, our CSFS framework is generalized to a rule-specific feature selection framework
On Some Issues Of Stochastic Comparisons & Their Applications
No full textOne of the important objectives of statistics is the comparison of random quantities. These comparisons are mainly based on the comparison of some measures associated with these random quantities. For example, it is very common to compare two random variables in terms of their means, medians, or variances. In some situations, comparisons based only on two single measures are not very informative. The necessity of providing more detailed comparisons of two random quantities has motivated the development of the theory of stochastic orders, which has grown significantly during the last 50 years. Stochastic order refers to comparing two random quantities in some stochastic sense. It is an important tool used in many diverse areas of statistics, reliability, economics, etc. Reliability theory and actuarial science are two most important areas where stochastic orders are studied extensively. Usual stochastic ordering, hazard rate ordering, reversed hazard rate ordering for lifetimes of series and parallel systems with heterogeneous and dependent componentshave beenestablished. Dispersive and star order for one heterogeneous and one homogeneous dependent series or parallel systems havealso been established. For two finite mixture models’ comparison, results have beenestablished under the usual stochastic order, hazard rate order and reversed hazard rate order. Various up and down-shifted ordering results have beenestablished for two important continuous mixture models: Frailty and Resilience. Data analysis has been done for illustration purposes. To find the optimal set of redundant components or systems, the usual stochastic ordering, hazard rate ordering and reversed hazard rate ordering of systems lifetimes under active redundancy allocation have beenestablished. Data analysis has also been donein this contextfor illustration purposes. Actuarial science is another area where stochastic ordershave extensivepotential for application. Usual stochastic ordering and star ordering results have beenestablished for the largest and aggregate claim amounts of two heterogeneous portfolios. Numerical exampleshave been provided in this context to illustrate the results thus obtained
Essays in the theory of allocation and voting
No full textThis thesis consists of two essays on allocation theory and one on voting theory. The first chapter analyses preference domains (called priority domains) where every strategy proof, non-bossy and neutral allocation rule is a priority rule. It considers two versions of neutrality: unanimous profile neutrality or UPN neutrality where the neutrality axiom applies only to preference profiles where all agents have a common preference ordering and full neutrality or FN neutrality, where the neutrality axiom applies generally. We show that a very simple condition characterises priority domains under the UPN axiom. If these domains satisfy a mild richness condition, they must be the universal domain. The class of priority domains under the FN axiom is larger than those satisfying only UPN. We identify an FN-priority domain that is of order 1 n relative to the universal domain. The second chapter analyses preference domains in voting environments where every strategy proof random social choice functions satisfying unanimity is a random dictatorships. We call these random-dictatorial domains. Pramanik (2015) identifies a class of domains called P-domains which are dictatorial i.e. every deterministic strategy proof social choice functions on these domains satisfying unanimity, is dictatorial. The main result of this chapter is that P-domain is random-dictatorial. A consequence of this result is that circular domains (Sato (2010)) are also random-dictatorial. The minimum size of a random-dictatorial domain satisfying minimal richness is shown to be twice the number of alternatives. This is the same as the corresponding lower bound for dictatorial domains. Our result stands in contrast to those in Chatterji et al. (2014) who showed that linked domains are not random-dictatorial. Linked domains were shown to be dictatorial in Aswal et al. (2003). The third chapter attempts to provide a justification of the non-bossiness axiom which is pervasive in the allocation literature. It has been criticised by Thomson (2016) on the grounds that it cannot be defended by appealing to various strategic and normative criteria. We show that in some special cases, non-bossiness is a simplifying assumption that can be imposed without loss of generality by an expected welfare maximising planner in a symmetric environment. We consider the case of three objects and three agents with a planner whose goal is to maximise the expected sum of welfare with respect to a uniform prior. We show that for every strategy proof, neutral and efficient allocation rule, there exists a strategy-proof, neutral and non-bossy allocation rule which yields the same expected welfare. We conjecture that this is true for an arbitrary number of agents. For the general case, we are able to show an equivalence in terms of expected welfare for a special class of bossy allocation rules
Design and Analysis of Authenticated Encryption Modes
No full textThis thesis proposes and analyses the security of a few symmetric key modes. The first three of them are NAEAD modes, named Oribatida, ISAP+ and OCB+. Oribatida is lightweight, sponge-based, INT-RUP secure and achieves better than the default PRF security of a keyed sponge. ISAP+ is an instance of a generic EtHM involving a PRF and a hash, a generalisation of ISAP-type modes. The generic sponge hash of ISAP is replaced with a feed-forward variant of it in ISAP+, which results in better security. OCB+ uses OTBC-3 (a nonce-respecting BBB secure offset-based tweakable block-cipher) in an OCB-like mode to achieve BBB privacy. We conclude with a BBB secure NE mode named CENCPP*, which is a public permutation-based variant of the block-cipher-based mode CENC as well as a variable output length version of SoEM. All the relevant security proofs have been done using a method named Coefficients H Techniqu
A new braiding index to assess river regulation effects in multi-thread channels: Insights from a highly regulated Himalayan river
No full textRiver regulation by dams and embankments drastically reduce/alter flow, which affects the natural channel pattern. Existing braiding indices have seldom incorporated the effects of diurnal flow variations caused by hydropeaking, leading to over/underestimation of the braiding intensity. These indices consider only the visible wet channels, ignoring the existence of dry channels that are activated only episodically during phases of water release from hydropower dams. We have extracted the dry channels (those that are periodically wet) coursing across the channel belt of the highly regulated River Tista from Landsat images between 1977 and 2014, using Normalized Difference Wetness Index values. These were combined with existing wet channel widths and numbers to formulate the Regulated Braiding Index (RBi) for characterizing channel braiding in the Tista\u27s course over the Himalayan foothill plains. Overall, the widths and numbers of wet channels decreased by 63% and 25%, respectively, during the regulated years (2003 and 2014) as compared to pre-dam years (1977 and 1995) due to the collective operation of 14 upstream dams/barrages (having cumulative ~89 million m3 reservoir capacity), whose operations reduced the braiding intensity (eliciting lower RBi values). Further, the number of high braiding reaches decreased by half while low braiding stretches rose by 10% in comparison to the pre-dam period due to reduced/regulated flow. Comparative analysis of derived RBi values with three existing braiding indices revealed that RBi values consistently decreased near barrages, bridges, and within embankment-confined reaches, unlike the other indices, thus providing a better framework for assessing expected river regulation effects