Binary Logistic Regression
Binary logistic regression models the probability of a two-category outcome from one or more predictors. This tutorial explains what an odds ratio means, how RAISINS reports fit and significance, how to check every assumption, and how to run the whole analysis code-free… Read more …
Binary logistic regression is the standard tool for modelling a yes/no outcome: diseased or healthy, present or absent, success or failure. Instead of predicting the outcome directly, it models the probability of the event through the logistic (S-shaped) curve, and reports each predictor’s influence as an odds ratio, a single, interpretable number that tells you how much the odds of the event change per unit of that predictor. This tutorial starts from first principles, what odds and log-odds are, how to read an odds ratio, and what the p-values actually test, then walks through the complete workflow in RAISINS: preparing data, choosing the binary outcome and predictors, enter vs stepwise model building, the coefficient/odds-ratio and model-fit tables, the assumption checks and influence diagnostics, machine-learning validation with ROC and calibration, publication-ready plots, automatic interpretation, and the RA-One AI assistant, all without writing a single line of code.
1 What is binary logistic regression?
Suppose you survey 150 rice plots and record, for each, whether blast disease appeared (1 = diseased, 0 = healthy) along with the temperature, humidity, leaf wetness and the variety’s resistance score. You suspect these factors drive disease occurrence, but by how much, and which ones actually matter once the others are accounted for?
You cannot fit an ordinary regression line here: the outcome is not a quantity but a category with two levels, and a straight line would happily predict probabilities below 0 or above 1. Binary logistic regression solves this by modelling the probability of the event through the logistic (sigmoid) function, an S-shaped curve that always stays between 0 and 1. From the fitted model you get two things at once:
- Prediction, given a plot’s temperature, humidity and so on, what is the probability it develops blast?
- Effect estimation, how much do the odds of disease change for a one-unit change in each predictor, holding the others fixed?
The model is fitted by maximum likelihood: of all possible curves, RAISINS picks the coefficient values that make your observed pattern of 0s and 1s most probable.
Binary logistic regression fits an S-shaped probability curve to your yes/no data and reports, for each predictor, an odds ratio, how strongly it raises or lowers the odds of the event, and whether that effect is larger than random noise would produce.
1.1 Odds, log-odds and the odds ratio
The odds of an event are the probability it happens divided by the probability it does not: a probability of 0.75 corresponds to odds of 3 (three to one). Logistic regression is linear on the log-odds scale:
\[\log\!\left(\frac{p}{1-p}\right) = \beta_0 + \beta_1 X_1 + \beta_2 X_2 + \cdots + \beta_k X_k\]
Every predictor gets a coefficient (β) on that log-odds scale, and exponentiating it gives the far friendlier odds ratio, OR = eβ, a partial effect: the multiplicative change in the odds of the event for a one-unit increase in that predictor while every other predictor is held constant.
- OR > 1, the predictor raises the odds (OR = 1.37 means +37% odds per unit).
- OR < 1, the predictor lowers the odds (OR = 0.66 means −34% odds per unit).
- OR = 1, no association.
For a categorical predictor, the odds ratio compares each level against that variable’s reference level: an OR of 2.0 for “Variety B” reads as twice the odds of disease compared with the reference variety, holding everything else constant. The intercept (β₀) is the model’s baseline, the log-odds of the event when every predictor sits at zero (or at its reference level).
1.2 The p-value, in plain language
For each coefficient RAISINS runs a test of the claim “this predictor has no effect” (β = 0, i.e. OR = 1) and reports a p-value, which answers:
If this predictor truly had no effect, how often would I see an odds ratio this far from 1 purely by chance?
A p-value of 0.40 means “quite often, nothing to see here.” A p-value below 0.001 means “almost never, this predictor is doing real work.” Researchers conventionally treat p < 0.05 as evidence that a predictor genuinely matters.
A p-value is not the probability that the effect is real, nor its size. A predictor can be highly significant yet practically trivial (OR = 1.02), or important yet non-significant in a small dataset. Always read the odds ratio (the size) and its 95% confidence interval next to the p-value, if the interval contains 1, the effect is not established at the 5% level.
1.3 The null and alternative hypotheses
Logistic regression tests each coefficient against a starting assumption that nothing is happening:
- H₀ : βⱼ = 0 (ORⱼ = 1), predictor j has no effect on the odds of the event.
- H₁ : βⱼ ≠ 0 (ORⱼ ≠ 1), predictor j does affect the odds.
A small p-value makes H₀ hard to believe, so you reject it and call the predictor significant. A large p-value means the data are compatible with “no effect,” so you fail to reject H₀, which is not the same as proving the predictor is useless, only that this dataset did not show its effect.
A little history: from the linear probability problem to Cox's logistic model
-
The origins of logistic regression trace back to Sir David Cox, who formally presented the model in 1958 in his landmark paper “The Regression Analysis of Binary Sequences” in the Journal of the Royal Statistical Society. Before Cox, researchers relied on linear probability models, which happily produced predicted probabilities below 0 or above 1. Cox resolved this by linking the linear predictor to the outcome probability through the logistic function, guaranteeing every prediction stays in [0, 1]. The logistic curve itself is older still, Pierre-François Verhulst introduced it in the 1830s to describe population growth, and Joseph Berkson championed the “logit” in bioassay in the 1940s. Over subsequent decades binary logistic regression became one of the most extensively applied statistical methods in medicine, epidemiology, agriculture, the social sciences and machine learning, prized for its robustness and the natural meaning of its coefficients as odds ratios. Its relevance has only grown in the era of data science, where binary classification problems are everywhere.
2 One predictor or several?
RAISINS fits both, and the only thing that changes is how many predictors you select.
| Model | Predictors | Typical question |
|---|---|---|
| Simple logistic regression | One predictor (X) | How do the odds of blast change with humidity alone? |
| Multiple logistic regression | Two or more predictors | What is the effect of humidity after accounting for temperature, leaf wetness and variety resistance? |
The power of the multiple model is that each odds ratio is a partial effect: it separates the contribution of one predictor from all the others. That power comes with one predictor-specific hazard: multicollinearity. When two predictors carry nearly the same information (say, two highly correlated weather measures), the model cannot decide which deserves the credit, and their individual odds ratios become unstable, even though the model’s overall predictions stay fine. RAISINS flags this automatically with the Variance Inflation Factor (VIF) in the Assumptions tab.
A predictor need not be numeric. A factor such as variety or soil type is entered as plain text labels; RAISINS automatically dummy-codes it, choosing one level as the reference and reporting each remaining level as an odds ratio against it. After you run the analysis, a Factor Settings panel appears above the results where you can change the reference level of any categorical predictor, choose the level that makes your odds ratios easiest to explain.
3 Model building: Enter vs Stepwise
The Analysis sidebar offers two ways to build the model under Select Regression Type:
| Method | What it does | When to reach for it |
|---|---|---|
| Simple Logistic (Enter) | Fits the model with every predictor you selected | You chose the predictors on scientific grounds and want them all reported |
| Step Wise Logistic (AIC) | Starts from the full model and drops terms to minimise AIC (via backward elimination) | You have many candidate predictors and want a parsimonious model |
When the stepwise method is used, RAISINS reports the AIC trace (which terms were dropped at each step and how AIC improved) and states the final selected model, and every downstream table refers to that final model. Remember that p-values computed after automatic selection are optimistic, because the selection itself is not accounted for; treat them as descriptive.
3.1 Machine-learning mode (train/test validation)
Separate from the method choice, the Analysis sidebar carries an Enable Machine Learning Approach switch. Turn it on and RAISINS randomly splits your data into a training set (default 80%, adjustable) and a held-out test set, fits the logistic model on the training portion, then evaluates how well it classifies the unseen test rows at the standard 0.5 probability threshold. The results then add a full classification report: the confusion matrix, accuracy, sensitivity (recall), specificity, precision and F1 score, and the Plots tab unlocks the ROC curve with AUC, the calibration plot and the predicted-probability distribution. The split uses a fixed random seed so the result is reproducible. Leave the switch off to fit the model on the entire dataset in the usual way.
The AIC, deviance and pseudo R² statistics describe fit on the data used to build the model; the machine-learning metrics describe classification of rows the model never saw. A model can fit its training data beautifully and still classify new plots poorly, the test-set metrics are the honest check against that overfitting.
4 Assumptions of logistic regression
Logistic regression is refreshingly free of the residual-normality and constant-variance assumptions of ordinary regression, but it has its own checklist. RAISINS provides a formal check for each item, so you never have to take the model on faith.
| Assumption | What it means | RAISINS check | What if it fails? |
|---|---|---|---|
| Binary outcome | The response has exactly two categories | View Data health check | Use multinomial/ordinal models for 3+ categories |
| Independence | Observations are unrelated to one another | Design-level question | Clustered/repeated data need mixed or GEE models |
| No multicollinearity | Predictors are not near-duplicates | VIF table | Drop or combine redundant predictors |
| Linearity of the logit | Each continuous predictor is linearly related to the log-odds | Box–Tidwell check | Transform, add a polynomial term, or categorise |
| Adequate calibration | Predicted probabilities match observed outcomes | Hosmer–Lemeshow test | Add predictors, interactions, or transform |
| Sufficient events | Roughly ≥ 10 events per predictor | Outcome distribution table | Reduce predictors or gather more data |
Separation is the dramatic one: if a predictor perfectly divides events from non-events, its coefficient explodes and its standard error becomes enormous, the model literally cannot estimate it. Non-linearity of the logit biases the odds ratios of the affected predictor. Multicollinearity never harms prediction; it only muddies the interpretation of individual odds ratios. Too few events makes everything unstable: with 5 events and 4 predictors, no software can save the analysis.
5 Getting to the module
Visit the RAISINS home page at www.raisins.live and open Data Analysis. Select the Binary Logistic Regression module (Figure 1). No programming is required: upload your data, choose the binary outcome and predictor columns, and RAISINS handles everything from computation to publication-ready output.
-02.png)
5.1 Citing RAISINS, and the record behind this module
To cite the platform in a paper, thesis or report, use the RAISINS citation (APA, Harvard and BibTeX) at www.raisins.live/citation.html. That is the primary reference.
Alongside it, each module has a CPRR, a Computational Provenance & Reproducibility Record, which answers the reviewer’s fair question, what exactly did the software do to my data? It states the R version and the exact version of every package, names the specific function behind each result (stats::glm() with binomial(link = "logit") for the model fit, MASS::stepAIC() for stepwise selection, car::vif() for multicollinearity, a standard decile-of-risk Hosmer–Lemeshow construction for goodness of fit, a Box–Tidwell refit for linearity of the logit, broom::augment() for the influence diagnostics, and pROC::roc() for the ROC curve and AUC), lists every default and decision rule (the 0.5 classification threshold, the fixed random seed for the train/test split, the influence cut-offs), and includes runnable R code reproducing each step on a public dataset. It carries its own DOI.
Cite the RAISINS paper as your primary reference for the platform, and add the CPRR as supporting documentation when a journal asks for computing details or when you want a precise methods section. The record supports the citation; it does not replace it.
6 Preview mode and Quick Tour
Before subscribing, explore the whole module using Preview mode from the Welcome page. It loads built-in datasets so you can try every feature, the analysis, assumption checks, plots and the RA-One assistant, without uploading your own data. First-time users are greeted with a Quick Tour: an interactive, step-by-step walkthrough that highlights each tab and control and explains what it does. You can replay it any time from the Quick Tour tab in the top navigation.
7 A working example
Throughout this tutorial we use an agricultural dataset of 150 rice plots. The response variable (Y) is Disease_Status (0 = Healthy, 1 = Diseased, blast disease), and the four predictors are Temp (°C), Humidity (%), Leaf_Wetness (hours/day) and Variety_Resistance (a resistance score). The goal is to learn which of these factors drive blast occurrence, and by how much the odds of disease change with each. The data layout is shown in Figure 3.
.png)
8 How to prepare your data
Your analysis is only as good as your data. Feed RAISINS clean data and it delivers powerful insight; feed it messy data and the results cannot be trusted. You have four routes:
- Create your dataset in MS Excel
- Build it directly within the RAISINS app (Create Data)
- Use a built-in Model dataset as a reference
- Create it through the RA-One chat assistant
9 Preparing data in MS Excel
Open a new blank Excel workbook with a single sheet and no stray content. Use a column-based layout: one column for the binary response variable (Y) and one column for each predictor (X), with one row per observation. The response column must contain exactly two distinct values, coded 0/1 or as two clearly named categories (e.g., “Healthy” and “Diseased”) with consistent spelling and capitalisation. Numeric predictors must be purely numeric; categorical predictors are entered as consistent text labels and RAISINS dummy-codes them automatically (you pick the reference level later in Factor Settings). Save as CSV, XLS or XLSX; CSV is recommended as it is lighter and loads faster. Avoid blank rows above the data and stray spaces in column names. For reference, see Figure 4.
Dataset creation rules
- Column naming, no spaces; use underscores (
_) or dots (.); avoid symbols such as %, #. Always start a column name with a letter. - Response variable, exactly two distinct values (0/1 or two labels); “Diseased” and “diseased” are treated as different categories, so keep capitalisation consistent.
- Data arrangement, start at the upper-left corner; the row above the data must not be blank.
- Cell management, do not type or delete in empty cells; if needed, select them, right-click and choose Clear Contents.
- Numeric vs categorical, measurement columns must be purely numeric; entries like “NA” or “–” cause errors. Categorical predictors should use consistent text labels with no trailing spaces.
- Missing values, avoid missing entries in the response or predictors; they cause observations to be dropped or the analysis to fail.
How to save as CSV in MS Excel
- Open your workbook, data on a single sheet, correctly arranged.
- File → Save As / Save a Copy, choose a location.
- Save as type → CSV (Comma delimited) (*.csv).
- Name the file without spaces (use underscores).
- Save.
💡 Tip: before saving, confirm the response column has exactly two distinct values, numeric columns are clean, and categorical labels are consistent.
10 Prepare using Create Data in RAISINS
Unsure about the format? RAISINS can build the template for you:
- Navigate to the Create Data tab
- Select the number of dependent (1) and independent variables
- Select the number of Observations (rows)
- Click Create
The layout appears as in Figure 5. Enter your values into the downloaded CSV (or paste them in), remembering that the response column must carry exactly two categories, then upload it under Analysis.
11 Download Model Datasets
To explore the module before using your own data, download a ready-made example:
- Navigate to the Datasets tab
- Click the Download CSV link for the dataset you want
- Use it as a formatting reference, or upload it directly to try the analysis
12 Creating a dataset using RA-One chat
RA-One, the built-in chat assistant, can create a correctly formatted dataset through a simple conversation. Open RA-One from its navigation tab or the floating chat bubble, tell it how many observations and how many predictors you need, and it generates an editable template in the required format, with a binary Y column ready for your 0/1 entries. Review it in the chat, download the CSV, and upload it under Analysis (Figure 7).
13 The Analysis tab
Figure 8 shows the Analysis tab. Click Browse in the sidebar to upload your CSV or Excel file. Selectors then appear: choose the model-building method under Select Regression Type (Simple Logistic or Step Wise Logistic, Section 3), pick the binary outcome under Dependent variable, and choose one or more predictors under Independent variable(s). Optionally flip on the Machine Learning switch and set the training percentage. Click Submit and every output appears instantly across the sub-tabs: Analysis Results, Assumptions, Plots & Graphs, Interpretation, FAQs and View Data. On the results panel you can also set the number of decimal digits used in the tables, and adjust the reference level of any categorical predictor through the Factor Settings panel.
.png)
14 Analysis results
The Analysis Results sub-tab presents the fitted model in stages: the outcome distribution (how many events and non-events, your events-per-variable check), the fitted logistic equation (in both logit and probability form, with a key mapping each x to its variable), the coefficient / odds-ratio table, and the model-fit statistics. A Download Report button exports the whole set as HTML, Word or PDF.
Table 1: Model Coefficients and Odds Ratios
Understanding the coefficient table
| Column | Meaning |
|---|---|
| Estimate (β) | Change in the log-odds of the event per one-unit rise in the predictor, holding the others fixed |
| Std. Error | Precision of the estimate; smaller is better |
| z / Wald statistic | Estimate ÷ Std. Error, tests whether β differs from zero |
| P Value | Probability of a statistic this extreme if β were truly zero |
| Odds Ratio (eβ) | Multiplicative change in the odds per unit of the predictor; the number you report |
| 95% CI for OR | Range for the true odds ratio; if it contains 1, the predictor is not significant at the 5% level |
| Significance | *** p < 0.001 · ** p < 0.01 · * p < 0.05 · (blank) not significant |
Interpretation from Figure 9
In the working example, Temp (OR = 1.37, p = 0.008) and Humidity (OR = 1.33, p = 0.014) are significant positive predictors: each 1 °C of temperature raises the odds of blast by about 37%, and each percentage point of humidity by about 33%, holding the other factors constant. Leaf_Wetness (OR = 1.22, p = 0.061) shows a positive trend that does not reach the 5% level. Variety_Resistance (OR = 0.66, p = 0.002) is a strongly protective factor, each unit of resistance score cuts the odds of disease by roughly 34%. The intercept is the log-odds of disease when every predictor is zero and rarely has a practical reading of its own. Each odds ratio is a partial effect: read it as “the change in the odds for a one-unit change in this predictor, with the others held constant.”
An odds ratio of 1.37 does not mean the disease is “37% more likely” in the probability sense, it means the odds are 37% higher. When the event is rare the two are close; when the event is common the odds ratio exaggerates the risk change. Report it as an odds ratio, which is exactly what it is.
Table 2: Model Fit Statistics
Understanding model fit statistics
| Statistic | Meaning |
|---|---|
| −2 Log-Likelihood (deviance) | Overall lack of fit; used to compare nested models |
| Likelihood-ratio test | Does the full model beat the intercept-only model? A small p-value says the predictors jointly matter |
| AIC | Deviance penalised for the number of parameters; lower is better, used by the stepwise method |
| Cox & Snell R² | Variance-explained analogue; cannot reach 1, so hard to read in absolute terms |
| Nagelkerke R² | Rescaled to 0–1; > 0.3 is generally moderate, > 0.5 strong |
| McFadden R² | Likelihood-based; values of 0.2–0.4 already indicate an excellent fit on this scale |
Interpretation from Figure 10
Here the likelihood-ratio test is highly significant (χ² = 38.72, df = 4, p < 0.001), so the four predictors jointly explain a real share of the variation in disease status. The Nagelkerke R² of 0.41 marks a moderate-to-strong model. Good fit statistics are encouraging, but they never certify that the model is correct, that is what the Assumptions tab is for.
Classification performance (machine-learning mode)
With the machine-learning switch on, the results add the held-out test-set report: the confusion matrix (predicted vs observed classes at the 0.5 threshold), accuracy, sensitivity/recall, specificity, precision and F1. In the working example the model classifies 83% of unseen plots correctly, catches 80% of truly diseased plots (sensitivity) and clears 86% of healthy ones (specificity), and its AUC of 0.88 (see the ROC plot in Section 16) confirms excellent discrimination.
The AUC (area under the ROC curve) summarises discrimination across all thresholds: 0.5 is coin-flipping, 0.7–0.8 acceptable, 0.8–0.9 excellent, > 0.9 outstanding. Always read it alongside accuracy, accuracy alone can flatter a model badly when one class dominates.
15 Assumptions
The Assumptions sub-tab gathers the formal checks that validate your model, each with a plain-language interpretation, and a “How do I read this tab?” guide at the top.
Multicollinearity, VIF
RAISINS reports the Variance Inflation Factor for each predictor. For a predictor \(X_j\),
\[VIF_j = \frac{1}{1 - R_j^2},\]
where \(R_j^2\) is the R² from regressing \(X_j\) on all other predictors. As a rule of thumb, VIF > 5 signals troublesome collinearity and VIF > 10 a serious problem. In the working example all values sit near 1, so multicollinearity is not a concern.
Goodness of fit, Hosmer–Lemeshow
The Hosmer–Lemeshow test groups observations into (about ten) risk groups by predicted probability and compares observed with expected event counts in each. A non-significant result (p ≥ 0.05) means the predicted probabilities agree with the observed outcomes, no evidence of poor fit. A significant result suggests the probabilities are mis-calibrated: consider missing predictors, interactions, or transforming a continuous predictor. With very large samples the test flags even trivial misfit, so read it together with the calibration plot.
Linearity of the logit, Box–Tidwell
For each continuous predictor, RAISINS refits the model with an extra \(x \cdot \log(x)\) term (the Box–Tidwell construction). A significant added term (p < 0.05) says that predictor’s relationship with the log-odds is probably not linear; the remedies are a transformation, a polynomial term, or sensible categorisation. Predictors containing zeros or negative values are reported as “not testable”, the check requires positive values.
Influence & Outlier Diagnostics
The tab closes with an Influence & Outlier Diagnostics table (Figure 15). It separates three distinct kinds of unusual observation: an outlier (poorly predicted, |standardized deviance residual| > 2, extreme above 3), a high-leverage point (unusual in the predictor space, leverage > 2p/n), and an influential point (one whose removal would visibly change the odds ratios, Cook’s distance > 4/n). For each observation it lists the predicted probability, the standardized deviance residual, the leverage and Cook’s distance, and adds a plain-language Flag. The cut-offs used are printed above the table, and the whole table is downloadable as CSV.
A flagged point may be a data-entry error (fix or remove it) or a genuine, important extreme case (keep it). Never delete points just because they are flagged; investigate, and where it matters, report the model with and without the point.
16 Plots & Graphs
The Plots & Graphs tab holds the full visualisation suite. Plots are selected from a row of icon buttons; the active plot appears below with a gear (Plot Settings) button in its corner. Every plot is fully customizable, title, axis labels, colours, point size, theme and more, and every plot can be downloaded in PNG, JPEG, TIFF, PDF and SVG at a chosen size and resolution. When machine-learning mode is on, a second row of icons unlocks the classification plots (ROC, calibration, probability distribution and the confusion-matrix heatmap).
.png)
Hover over any thumbnail below to read what it shows.
Logistic probability curve, the S-shaped fit of event probability against a chosen predictor; the quickest read on the direction and strength of a relationship.
Odds ratio plot, each predictor’s odds ratio as a point with its 95% confidence interval around the reference line at 1; bars crossing 1 are non-significant. A compact summary of effect sizes and significance.
Importance plot, predictors ranked by the strength of their evidence (−log₁₀ p); the taller the bar, the stronger that predictor’s contribution to the model.
Residual plot (diagnostic), standardized residuals by observation, coloured by outcome; look for a patternless band, isolated extreme points deserve a check in the influence table.
Deviance residual histogram (diagnostic), the distribution of deviance residuals; a roughly symmetric shape around zero supports the fit, long tails point at poorly explained observations.
ROC curve (machine-learning mode), sensitivity against 1 − specificity across all thresholds, with the AUC printed on the plot; the further the curve bows above the diagonal, the better the discrimination.
Machine-learning mode also unlocks the calibration plot (binned predicted vs observed proportions hugging the 45° line for a well-calibrated model), the predicted-probability distribution (how cleanly the two classes separate around the 0.5 threshold), and the confusion-matrix heatmap.
17 Interpretation
The Interpretation sub-tab gives a clear, plain-language write-up of your results (Figure 23). Tick the confirmation checkbox, click Generate Interpretation, and RAISINS composes the narrative: the model summary (method, sample size, split), a per-variable reading of every odds ratio (“each unit of Temp raises the odds of disease by 37%…”), which predictors are significant and at what level, the assumption checks that passed or failed (VIF, Hosmer–Lemeshow, linearity of the logit), the classification metrics when machine-learning mode is on, and an honest, data-driven conclusion with the Nagelkerke R², finishing with the RAISINS references, ready to adapt into a methods-and-results paragraph. The text is revealed with a short typing animation; a Stop button reveals the full write-up at once, and a Copy button places the plain-text version on your clipboard for pasting into a manuscript.
18 RA-One Chat
RA-One is the built-in conversational assistant, reachable from its own navigation tab or the floating chat bubble. Ask questions in plain language and it answers using your analysis, never generic advice and never invented numbers; if a value is not available it says so. Every reply is in plain English, with no code.
RA-One draws on the full set of results the app produces, the coefficient table with odds ratios and confidence intervals, the AIC, deviance and likelihood-ratio test, the pseudo R² measures, the fitted equation, the VIF, Hosmer–Lemeshow and Box–Tidwell checks, the influence and outlier diagnostics, and the machine-learning test-set metrics (accuracy, sensitivity, specificity, precision, F1, AUC and the confusion matrix) when that mode is on. It can explain what an odds ratio, p-value or confidence interval means for your specific model, walk you through why a check passed or failed, and answer general concept questions (odds vs probability, reference levels, separation, events-per-variable) so you build understanding alongside your results.
The same chat window can also prepare your data, build a correctly formatted template (Section 12) or fetch a model dataset (Section 11), and, most usefully, generate plots on request. Ask for “the probability curve”, “the odds ratio plot”, “the ROC curve” or a “correlation heatmap” and RA-One renders the figure directly in the chat. Each chat plot comes with a plot-type toolbar to switch graphics and a Plot Settings panel to pick the predictor and restyle it, and a one-click high-resolution download.
In a single conversation RA-One can interpret your results, build a data template, fetch a model dataset, and draw customizable plots; so much of a routine logistic-regression session can happen without ever leaving the chat.
19 FAQs
A dedicated FAQs tab answers common questions, when to use logistic regression versus linear regression or a chi-squared test, how categorical predictors and reference levels are handled, how to read an odds ratio for a continuous versus a binary predictor, what to do when the model does not converge (often separation), and how to report results in line with journal guidelines. If you are ever unsure how something works, start here.
20 View data
View Data is the first line of defence for data integrity. On upload, RAISINS runs an automated Health Check that validates column types and formatting: it verifies the response column carries exactly two distinct values, flags non-numeric entries and missing values in the predictors, and warns when the number of observations is thin relative to the number of predictors (the ≥ 10 events-per-variable rule of thumb). Resolving these before you click Submit ensures every output rests on clean, correctly typed data.
21 Wrapping up
Binary logistic regression rests on one honest question: once the noise is accounted for, how much does each factor really move the odds of the event? Everything else, the likelihood-ratio test, the goodness-of-fit and linearity checks, the influence diagnostics, the ROC curve, exists to make sure that question is answered fairly. RAISINS automates the machinery so you can concentrate on what the answer means for your research.
If you get stuck at any point, RA-One is available 24 × 7, or write to us at [email protected].